Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Luc Van GoolSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works542 from public data
- SwinIR: Image Restoration Using Swin Transformer5,285
A strong baseline model SwinIR is proposed for image restoration based on the Swin Transformer that outperforms state-of-the-art methods on different tasks by up to 0.14∼0.45dB, while the total number of parameters can be reduced byUp to 67%.
- 4,242
- RePaint: Inpainting using Denoising Diffusion Probabilistic Models2,341
This work proposes RePaint: A Denoising Diffusion Probabilistic Model (DDPM) based inpainting approach that is applicable to even extreme masks and outperforms state-of-the-art Autoregressive, and GAN approaches for at least five out of six mask distributions.
- 1,955
- A+: Adjusted Anchored Neighborhood Regression for Fast Super-Resolution1,516
This work proposes A+, an improved variant of Anchored Neighborhood Regression, which combines the best qualities of ANR and SF and builds on the features and anchored regressors from ANR but instead of learning the regressors on the dictionary it uses the full training material, similar to SF.
- Learning Discriminative Model Prediction for Tracking1,436
An end-to-end tracking architecture, capable of fully exploiting both target and background appearance information for target model prediction, derived from a discriminative learning loss by designing a dedicated optimization process that is capable of predicting a powerful model in only a few iterations.
- 1,351
- Plug-and-Play Image Restoration With Deep Denoiser Prior1,296
Experimental results demonstrate that the proposed plug-and-play image restoration with deep denoiser prior not only significantly outperforms other state-of-the-art model-based methods but also achieves competitive or even superior performance against state- of theart learning- based methods.
- Designing a Practical Degradation Model for Deep Blind Image Super-Resolution1,035
The experimental results demonstrate that the new degradation model can help to significantly improve the practicability of deep super-resolvers, thus providing a powerful alternative solution for real SISR applications.
- CDDFuse: Correlation-Driven Dual-Branch Feature Decomposition for Multi-Modality Image Fusion990
A novel Correlation-Driven feature Decomposition Fusion (CDDFuse) network that achieves promising results in multiple fusion tasks, including infrared- visible image fusion and medical image fusion, and can boost the performance in downstream infrared-visible semantic segmentation and object detection in a unified benchmark.
- Deep Expectation of Real and Apparent Age from a Single Image Without Facial Landmarks959
A deep learning solution to age estimation from a single face image without the use of facial landmarks is proposed and the IMDB-WIKI dataset is introduced, the largest public dataset of face images with age and gender labels.
- Retinexformer: One-stage Retinex-based Transformer for Low-light Image Enhancement884
A simple yet principled One-stage Retinex-based Framework (ORF), designed an Illumination-Guided Transformer (IGT) that utilizes illumination representations to direct the modeling of non-local interactions of regions with different lighting conditions, and obtains the algorithm, Retinexformer.
- O-HAZE: A Dehazing Benchmark with Real Hazy and Haze-Free Outdoor Images782
The first outdoor scenes database composed of pairs of real hazy and corresponding haze-free images is introduced, and is used to compare a representative set of state-of-the-art dehazing techniques, using traditional image quality metrics such as PSNR, SSIM and CIEDE2000.
- 769
- Probabilistic Regression for Visual Tracking751
This work proposes a probabilistic regression formulation and applies it to tracking, which is capable of modeling label noise stemming from inaccurate annotations and ambiguities in the task and substantially improves the performance.
- Generative Adversarial Networks for Extreme Learned Image Compression686
If a semantic label map of the original image is available, the learned image compression system can fully synthesize unimportant regions in the decoded image such as streets and trees from the label map, proportionally reducing the storage cost.
- Deep Unfolding Network for Image Super-Resolution683
This paper proposes an end-to-end trainable unfolding network which leverages both learningbased methods and model-based methods to super-resolve blurry, noisy images for different scale factors via a single model, while maintaining the advantages of learning- based methods.
- 646
- 623
- DSLR-Quality Photos on Mobile Devices with Deep Convolutional Networks594
An end-to-end deep learning approach that bridges the gap by translating ordinary photos into DSLR-quality images by learning the translation function using a residual convolutional neural network that improves both color rendition and image sharpness.
- The 2018 PIRM Challenge on Perceptual Image Super-Resolution592
This paper reports on the 2018 PIRM challenge on perceptual super-resolution (SR), held in conjunction with the Perceptual Image Restoration and Manipulation (PIRM) workshop at ECCV 2018, and concludes with an analysis of the current trends in perceptual SR, as reflected from the leading submissions.
- Conditional Probability Models for Deep Image Compression548
This paper proposes a new technique to navigate the rate-distortion trade-off for an image compression auto-encoder by using a context model: A 3D-CNN which learns a conditional probability model of the latent distribution of the auto- Encoder.
- Denoising Diffusion Models for Plug-and-Play Image Restoration511
Experimental results on three representative IR tasks demonstrate that DiffPIR achieves state-of-the-art performance on both the FFHQ and ImageNet datasets in terms of reconstruction faithfulness and perceptual quality with no more than 100 NFEs.
- NH-HAZE: An Image Dehazing Benchmark with Non-Homogeneous Hazy and Haze-Free Images499
This work presents an objective assessment of several state- of-the-art single image dehazing methods that were evaluated using NH-HAZE dataset, a non-homogeneous realistic dataset with pairs of real hazy and corresponding haze-free images.
- Dense-Haze: A Benchmark for Image Dehazing with Dense-Haze and Haze-Free Images499
Dense-Haze dataset aims to push significantly the state-of-the-art in single-image dehazing by promoting robust methods for real and various hazy scenes and contains 33 pairs of real hazy and corresponding haze-free images of various outdoor scenes.
- 496
- I-HAZE: A Dehazing Benchmark with Real Hazy and Haze-Free Indoor Images482
A new dataset -named I-HAZE- that contains 35 image pairs of hazy and corresponding haze-free (ground-truth) indoor images that allows us to objectively compare the existing image dehazing techniques using traditional image quality metrics such as PSNR and SSIM.
- DDFM: Denoising Diffusion Model for Multi-Modality Image Fusion480
This work proposes a novel fusion algorithm based on the denoising diffusion probabilistic model (DDPM) that can generate high-quality fused images with natural image generative priors and cross-modality information from source images.
- LocalViT: Analyzing Locality in Vision Transformers460
The locality mechanism is systematically investigated by carefully designed controlled experiments and the same locality mechanism is successfully applied to vision transformers with different architecture designs, which shows the generalization of the locality concept.
- Seven Ways to Improve Example-Based Single Image Super Resolution452
The Improved A+ (IA) method sets new stateof-the-art results outperforming A+ by up to 0.9dB on average PSNR whilst maintaining a low time complexity.
- VRT: A Video Restoration Transformer448
Experimental results on five tasks, including video super-resolution, video deblurring, video denoising, video frame interpolation and space-time video super-resolution, demonstrate that VRT outperforms the state-of-the-art methods by large margins on fourteen benchmark datasets.
- SRFlow: Learning the Super-Resolution Space with Normalizing Flow444
The proposed SRFlow is a normalizing flow based super-resolution method capable of learning the conditional distribution of the output given the low-resolution input, and directly accounts for the ill-posed nature of the problem, and learns to predict diverse photo-realistic high-resolution images.
- 443
- Mask-guided Spectral-wise Transformer for Efficient Hyperspectral Image Reconstruction423
A novel framework, Mask-guided Spectral-wise Transformer (MST), that treats each spectral feature as a token and calculates self-attention along the spectral dimension and significantly outperforms state-of-the-art (SOTA) methods on simulation and real HSI datasets while requiring dramatically cheaper computational and memory costs.
- Know Your Surroundings: Exploiting Scene Information for Object Tracking402
This work proposes a novel tracking architecture which can utilize scene information as dense localized state vectors, which can encode, for example, if the local region is target, background, or distractor and combined with the appearance model output to localize the target.
- Fast Optical Flow Using Dense Inverse Search393
The Dense Inverse Search-based method (DIS) is the efficient search of correspondences inspired by the inverse compositional image alignment proposed by Baker and Matthews (2001, 2004), making DIS ideal for real-time applications.
- Hough Transform and 3D SURF for Robust Three Dimensional Classification389
A new robust 3D shape classification method is proposed, which extends a robust 2D feature descriptor, SURF, to be used in the context of 3D shapes and shows how3D shape class recognition can be improved by probabilistic Hough transform based methods, already popular in 2D.
- AI Benchmark: Running Deep Neural Networks on Android Smartphones377
A study of the current state of deep learning in the Android ecosystem and describe available frameworks, programming models and the limitations of running AI on smartphones, as well as an overview of the hardware acceleration resources available on four main mobile chipset platforms.
- MST++: Multi-stage Spectral-wise Transformer for Efficient Spectral Reconstruction364
This work proposes a novel Transformer-based method, Multi-stage Spectral-wise Transformer (MST++), for efficient spectral reconstruction that significantly outperforms other state-of-the-art methods.
- 360
- 358
- Efficient and Explicit Modelling of Image Hierarchies for Image Restoration354
A new network architecture dubbed GRL is proposed to explicitly model image hierarchies in the Global, Regional, and Local range via anchored stripe self-attention, window self-Attention, and channel attention enhanced convolution.
- The Eighth Visual Object Tracking VOT2020 Challenge Results350
A significant novelty is introduction of a new VOT short-term tracking evaluation methodology, and introduction of segmentation ground truth in the VOT-ST2020 challenge – bounding boxes will no longer be used in theVDT challenges.
- 322
- 307
- Equivariant Multi-Modality Image Fusion299
This work proposes the Equivariant Multi-Modality imAge fusion (EMMA) paradigm, a novel training paradigm that encompasses a fusion module, a pseudo-sensing module, and an equivariant fusion module that enables the net training to follow the principles of the natural sensing-imaging process while satisfying the equivariant imaging prior.
- Frequency Separation for Real-World Super-Resolution282
DSGAN is proposed to introduce natural image characteristics in bicubic downscaled images to improve the performance of the SR model and separate the low and high image frequencies and treat them differently during training.
- Practical Blind Image Denoising via Swin-Conv-UNet and Data Synthesis270
Extensive experiments on AGWN removal and real image denoising demonstrate that the new network architecture design achieves state-of-the-art performance and the new degradation model can help to significantly improve the practicability.
- AI Benchmark: All About Deep Learning on Smartphones in 2019267
This paper evaluates the performance and compares the results of all chipsets from Qualcomm, HiSilicon, Samsung, MediaTek and Unisoc that are providing hardware acceleration for AI inference and discusses the recent changes in the Android ML pipeline.
- GLU-Net: Global-Local Universal Network for Dense Flow and Correspondences262
This work proposes a universal network architecture that is directly applicable to all the aforementioned dense correspondence problems, and achieves both high accuracy and robustness to large displacements by investigating the combined use of global and local correlation layers.
- Practical Full Resolution Learned Lossless Image Compression259
The first practical learned lossless image compression system, L3C, is proposed and it outperforms the popular engineered codecs, PNG, WebP and JPEG 2000, and finds that learning the auxiliary representation is crucial and outperforms predefined auxiliary representations such as an RGB pyramid significantly.
- Learning for Video Compression With Hierarchical Quality and Recurrent Enhancement253
The experiments validate that the HLVC approach advances the state-of-the-art of deep video compression methods, and outperforms the "Low-Delay P (LDP) very fast" mode of x265 in terms of both PSNR and MS-SSIM.
- HDNet: High-resolution Dual-domain Learning for Spectral Compressive Imaging250
A high-resolution dual-domain learning network (HDNet) for HSI reconstruction with HR pixel-level attention and frequency-level refinement that achieves SOTA performance on simulated and real HSI datasets.
- Video super-resolution based on deep learning: a comprehensive survey249
This work comprehensively investigates 37 state-of-the-art VSR methods based on deep learning and proposes a taxonomy and classify the methods into seven sub-categories according to the ways of utilizing inter-frame information.
- Night-to-Day Image Translation for Retrieval-based Localization248
This paper proposes ToDayGAN – a modified image-translation model to alter nighttime driving images to a more useful daytime representation, and improves localization performance by over 250% compared the current state-of-the-art, in the context of standard metrics in multiple categories.
- InstructIR: High-Quality Image Restoration Following Human Instructions234
This work presents the first approach that uses human-written instructions to guide the image restoration model, InstructIR, which can recover high-quality images from their degraded counterparts, considering multiple degradation types.
- Replacing Mobile Camera ISP with a Single Deep Learning Model223
PyNET is presented, a novel pyramidal CNN architecture designed for fine-grained image restoration that implicitly learns to perform all ISP steps such as image demosaicing, denoising, white balancing, color and contrast correction, demoireing, etc.
- Group Sparsity: The Hinge Between Filter Pruning and Decomposition for Network Compression220
This paper analyzes two popular network compression techniques, i.e. filter pruning and low-rank decomposition, in a unified sense and proposes to compress the whole network jointly instead of in a layer-wise manner.
- WESPE: Weakly Supervised Photo Enhancer for Digital Cameras219
This work introduces a weakly supervised photo enhancer (WESPE) - a novel image-to-image Generative Adversarial Network-based architecture that produces comparable or improved qualitative results with state-of-the-art strongly supervised methods.
- Fourier Space Losses for Efficient Perceptual Image Super-Resolution216
It is shown that it is possible to improve the performance of a recently introduced efficient generator architecture solely with the application of the proposed loss functions, and a discriminator architecture working directly in the Fourier domain to better match the target HF distribution is designed.
- ComboGAN: Unrestrained Scalability for Image Domain Translation213
This paper proposes a multi-component image translation model and training scheme which scales linearly - both in resource consumption and time required - with the number of domains and demonstrates its capabilities on a dataset of paintings by 14 different artists.
- 199
- Unsupervised Learning for Real-World Super-Resolution195
This work learns to invert the effects of bicubic downsampling in order to restore the natural image characteristics present in the data, and can be trained with direct pixel-wise supervision in the high resolution domain, while robustly generalizing to real input.
- NTIRE 2020 Challenge on Spectral Reconstruction from an RGB Image194
A new, larger-than-ever, natural hyperspectral image data set is presented, containing a total of 510 HS images, and the proposed methods gauge the state-of-the-art in spectral reconstruction from an RGB image.
- NTIRE 2020 Challenge on Real-World Image Super-Resolution: Methods and Results191
The NTIRE 2020 challenge addresses the real world setting, where paired true high and low-resolution images are unavailable, and the ultimate goal is to achieve the best perceptual quality, evaluated using a human study.
- 189
- Applying artificial intelligence to assess the impact of orthognathic treatment on facial attractiveness and estimated age183
It is illustrated that artificial intelligence might be considered to score facial attractiveness and apparent age in orthognathic patients.
- Learning What to Learn for Video Object Segmentation182
This work introduces an end-to-end trainable VOS architecture that integrates a differentiable few-shot learning module that is designed to predict a powerful parametric model of the target by minimizing a segmentation error in the first frame.
- 179
- Deep Burst Super-Resolution176
A novel architecture for the burst super-resolution task, which takes multiple noisy RAW images as input, and generates a denoised, super-resolved RGB image as output by explicitly aligning deep embeddings of the input frames using pixel-wise optical flow.
- Unsupervised Real-world Image Super Resolution via Domain-distance Aware Training173
This paper proposes a novel domain-distance aware super-resolution (DASR) approach for unsupervised real-world image SR and results show that DASR consistently outperforms state-of-the-art un supervised SR approaches in generating SR outputs with more realistic and natural textures.
- 173
- Efficient Video Super-Resolution through Recurrent Latent Space Propagation173
An efficient recurrent latent space propagation algorithm for fast VSR that introduces high-dimensional latent states to propagate temporal information between frames in an implicit manner and outperform current state-of-the-art method DUF with over 70x speed-up.
- Single-Model and Any-Modality for Video Object Tracking172
This work introduces Un-Track, a Unified Tracker of a single set of parameters for any modality, which surpasses both SOTA unified trackers and modality-specific counterparts, validating the effectiveness and practicality.
- Learning Accurate Dense Correspondences and When to Trust Them171
This work aims to estimate a dense flow field relating two images, coupled with a robust pixel-wise confidence map indicating the reliability and accuracy of the prediction, and develops a flexible probabilistic approach that jointly learns the flow prediction and its uncertainty.
- Learning for Video Compression With Recurrent Auto-Encoder and Recurrent Probability Model169
This paper proposes a Recurrent Learned Video Compression (RLVC) approach with the Recurrent Auto-Encoder (RAE) and Recurrent Probability Model (RPM), which achieves the state-of-the-art learned video compression performance in terms of both PSNR and MS-SSIM.
- 168
- Jointly Optimized Regressors for Image Super‐resolution166
A collection of regressors are jointly learned, which collectively yield the smallest super‐resolving error for all training data, and this method is conceptually simple and computationally efficient, yet very effective.
- Towards Flexible Blind JPEG Artifacts Removal162
This paper proposes a flexible blind convolutional neural network that can predict the adjustable quality factor to control the trade-off between artifacts removal and details preservation and achieves favorable performance against state-of-the-art methods in terms of both quantitative metrics and visual quality.
- PIRM Challenge on Perceptual Image Enhancement on Smartphones: Report161
This paper reviews the first challenge on efficient perceptual image enhancement with the focus on deploying deep learning models on smartphones and proposes solutions that significantly improved baseline results defining the state-of-the-art for image enhancement on smartphones.
- Flow-based Kernel Prior with Application to Blind Super-Resolution160
This paper proposes a normalizing flow-based kernel prior (FKP) that optimizes the kernel in the latent space rather than the network parameter space, which allows it to generate reasonable kernel initialization, traverse the learned kernel manifold and improve the optimization stability.
- 155
- 149
- Source-free Depth for Object Pop-out142
This work adapts depth inference models for object segmentation using the objects’ "pop-out" prior in 3D to adapt the inferred depth maps such that objects can be localized using only 3D information.
- DHP: Differentiable Meta Pruning via HyperNetworks140
This paper introduces a differentiable pruning method via hypernetworks for automatic network pruning, and extensive experiments are conducted on various networks for image classification, single image super-resolution, and denoising.
- 140
- PDC-Net+: Enhanced Probabilistic Dense Correspondence Network135
The Enhanced Probabilistic Dense Correspondence Network, PDC-Net+, is proposed, capable of estimating accurate dense correspondences along with a reliable confidence map, and a flexible probabilistic approach that jointly learns the flow prediction and its uncertainty is developed.
- Adversarial Sampling for Active Learning135
This paper proposes ASAL, a new GAN based active learning method that generates high entropy samples that outperforms similar methods and clearly exceeds the established baseline (random sampling).
- NTIRE 2018 Challenge on Spectral Reconstruction from RGB Images135
This paper reviews the first challenge on spectral image reconstruction from RGB images, i.e., the recovery of whole-scene hyperspectral (HS) information from a 3-channel RGB image.
- Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescaling131
The hierarchical conditional flow (HCFlow) is proposed as a unified framework for image SR and image rescaling and achieves state-of-the-art performance in terms of both quantitative metrics and visual quality.
- 130
- Complexity Experts are Task-Discriminative Learners for Any Image Restoration128
The proposed MoCE-IR model outperforms state-of-the-art methods, affirming its efficiency and practical applicability and introducing “complexity experts” – flexible expert blocks with varying computational complexity and receptive fields.
- The Ninth Visual Object Tracking VOT2021 Challenge Results127
The Visual Object Tracking challenge VOT2021 is the ninth annual tracker benchmarking activity organized by the VOT initiative; results of 71 trackers are presented; many are state-of-the-art trackers published at major computer vision conferences or in journals in recent years.
- NTIRE 2020 Challenge on NonHomogeneous Dehazing125
This paper reviews the NTIRE 2020 Challenge on Non-Homogeneous Dehazing of images (restoration of rich details in hazy image) and proposed solutions gauge the state-of-the-art in image dehazing.
- CiaoSR: Continuous Implicit Attention-in-Attention Network for Arbitrary-Scale Image Super-Resolution123
This paper explicitly design an implicit attention network to learn the ensemble weights for the nearby local features in a continuous implicit attention-in-attention network, called CiaoSR, which significantly outperforms the existing single image SR methods with the same backbone.
- 123
- Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution120
Extensive experiments show that the proposed MANet not only performs favorably for both spatially variant and invariant kernel estimation, but also leads to state-of-the-art blind SR performance when combined with non-blind SR methods.
- Apparent and Real Age Estimation in Still Images with Deep Residual Regressors on Appa-Real Database120
- Revisiting Random Channel Pruning for Neural Network Compression119
The proposed approach provides a new way to compare different methods, namely how well they behave compared with random pruning, and it is shown that this simple strategy works quite well compared with other channel pruning methods.
- 117
- AIM 2019 Challenge on Real-World Image Super-Resolution: Methods and Results110
The AIM 2019 challenge on real world super-resolution addresses the real world setting, where paired true high and low-resolution images are unavailable, and aims to advance the state-of-the-art and provide a standard benchmark for this newly emerging task.
- 110
- Exemplar Guided Face Image Super-Resolution Without Facial Landmarks110
A convolutional neural network (CNN)-based solution, namely GWAInet, which applies super-resolution by a factor 8x on face images guided by another unconstrained HR face image of the same person with possible differences in age, expression, pose or size is proposed.
- Learning Filter Basis for Convolutional Neural Network Compression108
This paper tries to reduce the number of parameters of CNNs by learning a basis of the filters in convolutional layers, and validate the proposed solution for multiple CNN architectures on image classification and image super-resolution benchmarks.
- NTIRE 2021 Challenge on Image Deblurring106
The winning methods demonstrate the state-of-the-art performance on the image deblurring task with the jointly combined artifacts.
- NTIRE 2020 Challenge on Real Image Denoising: Dataset, Methods and Results106
This paper reviews the NTIRE 2020 challenge on real image denoising with focus on the newly introduced dataset, the proposed methods and their results, based on the SIDD benchmark.
- 102
- AIM 2020 Challenge on Efficient Super-Resolution: Methods and Results101
The AIM 2020 challenge on efficient single image super-resolution was to super-resolve an input image with a magnification factor x4 based on a set of prior examples of low and corresponding high resolution images with focus on the proposed solutions and results.
- Rendering Natural Camera Bokeh Effect with Deep Learning100
This paper presents a large-scale bokeh dataset consisting of 5K shallow / wide depth-of-field image pairs captured using the Canon 7D DSLR with 50mm f/1.8 lenses, and proposes to learn a realistic shallow focus technique directly from the photos produced by DSLR cameras.
- 99
- NTIRE 2022 Challenge on Efficient Super-Resolution: Methods and Results94
The NTIRE 2022 challenge was to super-resolve an input image with a magnification factor of ×4 based on pairs of low and corresponding high resolution images and the aim was to design a network for single image super-resolution that achieved improvement of efficiency measured according to several metrics.
- 93
- The Ninth NTIRE 2024 Efficient Super-Resolution Challenge Report91
This paper provides a comprehensive review of the NTIRE 2024 challenge, focusing on efficient single-image super-resolution (ESR) solutions and their outcomes, and gauge the state-of-the-art in efficient single-image super-resolution.
- AIM 2020 Challenge on Learned Image Signal Processing Pipeline90
This paper reviews the second AIM learned ISP challenge and provides the description of the proposed solutions and results, defining the state-of-the-art for practical image signal processing pipeline modeling.
- VA-DepthNet: A Variational Approach to Single Image Depth Prediction89
It is shown that imposing first-order variational constraints in the scene space together with popular encoder-decoder-based network architecture design provides excellent results for the supervised SIDP task.
- Some Like It Hot — Visual Guidance for Preference Prediction89
This work improves the state of-the-art, but also predicts - based on someone's known preferences - how much that particular person is attracted to a novel face, and validates the collaborative filtering solution on the standard MovieLens rating dataset.
- NTIRE 2021 Challenge on High Dynamic Range Imaging: Dataset, Methods and Results87
This paper reviews the first challenge on high-dynamic range (HDR) imaging that was part of the New Trends in Image Restoration and Enhancement (NTIRE) workshop, held in conjunction with CVPR 2021.
- NTIRE 2024 Challenge on Low Light Image Enhancement: Methods and Results85
This paper meticulously evaluates the state-of-the-art advancements in enhancing low-light images, reflecting the significant progress and creativity in this field.
- NTIRE 2022 Challenge on Perceptual Image Quality Assessment84
This paper reports on the NTIRE 2022 challenge on perceptual image quality assessment (IQA), held in conjunction with the New Trends in Image Restoration and Enhancement workshop (NTIRE) workshop at CVPR 2022 to address the emerging challenge of IQA by perceptual image processing algorithms.
- SESAME: Semantic Editing of Scenes by Adding, Manipulating or Erasing Objects84
SESAME, a novel generator-discriminator pair for Semantic Editing of Scenes by Adding, Manipulating or Erasing objects, is proposed, composed of two input streams that independently process the image and its semantics, using the latter to manipulate the results of the former.
- NTIRE 2024 Challenge on Image Super-Resolution (×4): Methods and Results83
This paper reviews the NTIRE 2024 challenge on image super-resolution (×4), highlighting the solutions proposed and the outcomes obtained, and offers a comprehensive overview of current trends in this field.
- Adaptive and Weighted Collaborative Representations for image classification82
The Weighted Collaborative Representation Classifier (WCRC) improves the classification performance over that of the original formulation, while keeping the simplicity and the speed of the originally CRC-RLS formulation.
- Deep Reparametrization of Multi-Frame Super-Resolution and Denoising80
The deep reparametrization of the maximum a posteriori formulation commonly employed in multi-frame image restoration tasks is proposed, derived by introducing a learned error metric and a latent representation of the target image, which transforms the MAP objective to a deep feature space.
- Real-Time Quantized Image Super-Resolution on Mobile NPUs, Mobile AI 2021 Challenge: Report76
The first Mobile AI challenge is introduced, where the target is to develop an end-to-end deep learning-based image super-resolution solutions that can demonstrate a real-time performance on mobile or edge NPUs.
- Facial attractiveness of cleft patients: a direct comparison between artificial-intelligence-based scoring and conventional rater groups76
AI could become a helpful tool to describe facial attractiveness, but the present results indicate that important adjustments are needed on AI models, to improve the interpretation of the impact of cleft features on facial attractiveness.
- 72
- Outdoor Plant Segmentation With Deep Learning for High-Throughput Field Phenotyping on a Diverse Wheat Dataset70
The results show that a small but carefully chosen and annotated set of images can provide a good basis for a powerful segmentation pipeline, and the proposed method achieves better performance on the selected dataset in spite of using a deep learning approach with limited data.
- Learned Smartphone ISP on Mobile NPUs with Deep Learning, Mobile AI 2021 Challenge: Report70
The target was to develop an end-to-end deep learning-based image signal processing (ISP) pipeline that can replace classical hand-crafted ISPs and achieve nearly real-time performance on smartphone NPUs.
- Real-Time Video Super-Resolution on Smartphones with Deep Learning, Mobile AI 2021 Challenge: Report69
The target is to develop an end-to-end deep learning-based video super-resolution solutions that can achieve a realtime performance on mobile GPUs and can upscale videos to HD resolution at up to 80 FPS while demonstrating high fidelity results.
- 68
- 68
- Sparse Representation Based Projections68
The results show that preserving the sparse representation of the signals from the original space in the (lower) dimensional projected space is beneficial for several benchmarks and corroborate the usefulness of the proposed sparse representation based linear and non-linear projections.
- Perceptual Learned Video Compression with Recurrent Conditional GAN67
The experimental results show that the learned PLVC model compresses video with good perceptual quality at low bit-rate, and that it outperforms the official HEVC test model (HM 16.20) and the existing learned video compression approaches for several perceptual quality metrics and user studies.
- SMIT: Stochastic Multi-Label Image-to-Image Translation67
This work proposes a joint framework of diversity and multi-mapping image-to-image translations, using a single generator to conditionally produce countless and unique fake images that hold the underlying characteristics of the source image.
- Logo Synthesis and Manipulation with Clustered Generative Adversarial Networks67
This paper builds a dataset - LLD - of 600k+ logos crawled from the world wide web and proposes the use of synthetic labels obtained through clustering to disentangle and stabilize GAN training, and validate this approach on CIFAR-10 and ImageNet-small to demonstrate its generality.
- 67
- SiNeRF: Sinusoidal Neural Radiance Fields for Joint Pose Estimation and Scene Reconstruction65
Sinusoidal Neural Radiance Fields (SiNeRF) is proposed that leverage sinusoidal activations for radiance mapping and a novel Mixed Region Sampling (MRS) for selecting ray batch efficiently and achieves comprehensive significant improvements in image synthesis quality and pose estimation accuracy.
- 64
- Spherical Space Feature Decomposition for Guided Depth Map Super-Resolution64
The Spherical Space feature Decomposition Network (SSDNet) is proposed to solve the above issues and can achieve state-of-the-art results on four test datasets, as well as generalize to real-world scenes.
- AIM 2019 Challenge on Constrained Super-Resolution: Methods and Results63
This paper reviews the AIM 2019 challenge on constrained example-based single image super-resolution with focus on proposed solutions and results and gauges the state-of-the-art in single imagesuper-resolution.
- A Brief Review of Image Denoising Algorithms and Beyond61
An overview of prior modeling approaches used in image denoising task is presented, and conventional sparse representation based Denoising algorithms, low-rank based denoizing algorithms and recently proposed deep neural networks based approaches are reviewed.
- 60
- 60
- 59
- 59
- 59
- Audio-Visual Efficient Conformer for Robust Speech Recognition58
This work proposes to improve the noise robustness of the recently proposed Efficient Conformer Connectionist Temporal Classification (CTC)-based architecture by processing both audio and visual modalities and replaces the E efficient Conformer grouped attention by a more efficient and simpler attention mechanism that is called patch attention.
- Rip Current Segmentation: A Novel Benchmark and YOLOv8 Baseline Results57
A comprehensive dataset containing 2,466 images with newly created polygonal annotations for instance segmentation, used for training and validation and training a deep learning model for instance segmentsation of rip currents is introduced.
- AIM 2019 Challenge on Image Demoireing: Methods and Results57
This paper reviews the first-ever image demoireing challenge that was part of the Advances in Image Manipulation (AIM) workshop, held in conjunction with ICCV 2019, and focuses on the proposed solutions and their results.
- 56
- DeFlow: Learning Complex Image Degradations from Unpaired Data with Conditional Flows55
DeFlow, a method for learning stochastic image degradations from unpaired data based on a novel unpaired learning formulation for conditional normalizing flows, is proposed, which allows to learn the conditional distribution of a noisy image given the clean input by solely minimizing the negative log-likelihood of the marginal distributions.
- NTIRE 2021 Challenge on Perceptual Image Quality Assessment55
The NTIRE 2021 challenge on perceptual image quality assessment (IQA), held in conjunction with the New Trends in Image Restoration and Enhancement workshop (NTIRE) workshop at CVPR 2021, achieved much better results than existing IQA methods, while the winning method can demonstrate state-of-the-art performance.
- 55
- NTIRE 2024 Quality Assessment of AI-Generated Content Challenge54
This challenge is to address a major challenge in the field of image and video processing, namely, Image Quality Assessment (IQA) and Video Quality Assessment (VQA) for AI-Generated Content (AIGC).
- NTIRE 2020 Challenge on Perceptual Extreme Super-Resolution: Methods and Results54
This paper reviews the NTIRE 2020 challenge on perceptual extreme super-resolution with focus on proposed solutions and results and gauges the state-of-the-art in single image super- resolution.
- 53
- 53
- NTIRE 2023 Challenge on Light Field Image Super-Resolution: Dataset, Methods and Results52
The first NTIRE challenge on light field (LF) image super-resolution (SR), which aims at super-resolving LF images under the standard bicubic degradation with a magnification factor of 4, develops a new LF dataset called NTIRE-2023 for validation and test, and provides a toolbox called BasicLFSR to facilitate model development.
- 52
- 52
- AIM 2020 Challenge on Rendering Realistic Bokeh52
The proposed solutions significantly improved the baseline results, defining the state-of-the-art for practical bokeh effect rendering problem.
- See More Details: Efficient Image Super-Resolution by Experts Mining51
This work introduces SeemoRe, an efficient SR model employing expert mining, strategically incorporates experts at different levels, adopting a collaborative methodology, allowing the model to achieve an optimal performance with minimal computational costs in efficient settings.
- 51
- Fast Stixel Computation for Fast Pedestrian Detection51
This work proposes a method to estimate the ground-obstacles boundary (and its distance), without computing a depth map, by properly parametrizing the search space in the image plane, and improves the algorithmic performance.
- AIM 2020: Scene Relighting and Illumination Estimation Challenge50
The novel VIDIT dataset used in the AIM 2020 challenge and the different proposed solutions and final evaluation results over the 3 challenge tracks are presented.
- Predict, Prevent, and Evaluate: Disentangled Text-Driven Image Manipulation Empowered by Pre-Trained Vision-Language Model49
A novel framework, i.e., Predict, Prevent, and Evaluate (PPE), for disentangled text-driven image manipulation that requires little manual annotation while being applicable to a wide variety of ma-nipulations.
- 49
- 48
- 47
- Object Segmentation by Mining Cross-Modal Semantics47
A novel approach by mining the Cross-Modal Semantics to guide the fusion and decoding of multimodal features, with the aim of controlling the modal contribution based on relative entropy.
- 47
- Advancing Learned Video Compression With In-Loop Frame Prediction47
This paper proposes an Advanced Learned Video Compression (ALVC) approach with the in-loop frame prediction module, which is able to effectively predict the target frame from the previously compressed frames, without consuming any bit-rate.
- Towards Spectral Estimation from a Single RGB Image in the Wild46
This work uses a reference spectrum as provided by a hyperspectral image camera, and proposes efficient deep learning solutions for sensitivity function estimation and spectral reconstruction from a single RGB image.
- Soft-to-Hard Vector Quantization for End-to-End Learned Compression of Images and Neural Networks.46
This work presents a new approach to learn compressible representations in deep architectures with an end-to-end training strategy based on a soft (continuous) relaxation of quantization and entropy, which is anneal to their discrete counterparts throughout training.
- A Training-free Classification Framework for Textures, Writers, and Materials46
The idea of a training-free texture classification scheme, which requires no training, is advocated not only for traditional texture benchmarks, but also for the identification of materials and of the writers of musical scores.
- 46
- NTIRE 2022 Challenge on High Dynamic Range Imaging: Methods and Results45
This paper reviews the challenge on constrained high dynamic range (HDR) imaging that was part of the New Trends in Image Restoration and Enhancement (NTIRE) workshop, held in conjunction with CVPR 2022, and the competition set-up, datasets, proposed methods and their results.
- NTIRE 2020 Challenge on Image Demoireing: Methods and Results45
This paper reviews the Challenge on Image Demoireing that was part of the New Trends in Image Restoration and Enhancement (NTIRE) workshop, held in conjunction with CVPR 2020, and examines the current state-of-the-art in image and burst image demoireing problems.
- 45
- NTIRE 2025 Challenge on Cross-Domain Few-Shot Object Detection: Methods and Results44
An overview of the 1st NTIRE 2025 CD-FSOD Challenge is presented, highlighting the proposed solutions and summarizing the results submitted by the participants.
- How to Train Your Energy-Based Model for Regression44
A simple yet highly effective extension of noise contrastive estimation is proposed, and its performance is compared to six popular methods from literature on the tasks of 1D regression and object detection to suggest that its training method should be considered the go-to approach.
- 43
- NTIRE 2021 Challenge on Quality Enhancement of Compressed Video: Methods and Results43
This paper reviews the first NTIRE challenge on quality enhancement of compressed video, with a focus on the proposed methods and results, and gauge the state of the art of video quality enhancement.
- NTIRE 2021 Multi-modal Aerial View Object Classification Challenge43
The first Challenge on Multimodal Aerial View Object Classification (MAVOC) is introduced in conjunction with the NTIRE 2021 workshop at CVPR and shows significant improvement of more than 15% accuracy from the current baselines for each track of the competition.
- NTIRE 2021 Challenge on Burst Super-Resolution: Methods and Results43
The NTIRE2021 challenge on burst super-resolution was to generate a clean RGB image with 4 times higher resolution given a RAW noisy burst as input, and the top-performing methods set a new state-of-the-art for the burstsuper-resolution task.
- AIM 2020 Challenge on Real Image Super-Resolution: Methods and Results43
This paper introduces the real image Super-Resolution challenge that was part of the Advances in Image Manipulation (AIM) workshop, held in conjunction with ECCV 2020, and gauges the state-of-the-art approaches for real image SR in terms of PSNR and SSIM.
- CARN: Convolutional Anchored Regression Network for Fast and Accurate Single Image Super-Resolution43
The convolutional anchored regression network (CARN) is proposed for fast and accurate single image super-resolution (SISR), Inspired by locally linear regression methods (A+ and ARN), and achieves the best speed and accuracy trade-off among the SR methods.
- 42
- 42
- Local Memory Attention for Fast Video Semantic Segmentation42
A novel neural network module is proposed that transforms an existing single-frame semantic segmentation model into a video semantic segmentsation pipeline and integrates into two popular semantic segmentations networks: ERFNet and PSPNet.
- NTIRE 2022 Challenge on Super-Resolution and Quality Enhancement of Compressed Video: Dataset, Methods and Results41
This paper reviews the NTIRE 2022 Challenge on Super-Resolution and Quality Enhancement of Compressed Video and proposed the LDV 2.0 dataset, which includes theLDV dataset (240 videos) and 95 additional videos.
- 41
- 41
- 40
- 40
- Perceptual Image Enhancement for Smartphone Real-Time Applications40
This paper proposes LPIENet, a lightweight network for perceptual image enhancement, with the focus on deploying it on smartphones, and shows that, with much fewer parameters and operations, the model can deal with the mentioned artifacts and achieve competitive performance compared with state-of-the-art methods on standard benchmarks.
- Efficient Video Semantic Segmentation with Labels Propagation and Refinement40
The proposed Efficient Video Segmentation (EVS) pipeline achieves accuracy levels competitive to the existing real-time methods for semantic image segmentation (mIoU above 60%), while achieving much higher frame rates.
- AIM 2019 Challenge on Image Demoireing: Dataset and Study40
This paper introduces a novel dataset, called LCDMoire, which was created for the first-ever image demoireing challenge that was part of the Advances in Image Manipulation (AIM) workshop, held in conjunction with ICCV 2019.
- XTrack: Multimodal Training Boosts RGB-X Video Object Trackers39
This work proposes a mixture-of-experts framework designed for multimodal video object tracking, and implements a classifier with weak loss tasked with distinguishing between modalities, which shows the benefit of the proposed method for RGB-X tracker during inference, with an average precision improvement over the current SOTA.
- 39
- The Tenth NTIRE 2025 Efficient Super-Resolution Challenge Report39
This paper presents a comprehensive review of the NTIRE 2025 Challenge on Single-Image Efficient Super-Resolution, emphasizing groundbreaking advancements in state-of-the-art single-image ESR techniques.
- 39
- Graph Transformer GANs for Graph-Constrained House Generation39
A novel graph Transformer generative adversarial network (GTGAN) to learn effective graph node relations in an end-to-end fashion for the challenging graph-constrained house generation task and a new node classification-based discriminator to preserve the high-level semantic and discriminative node features for different house components.
- Fast and Accurate Single-Image Depth Estimation on Mobile Devices, Mobile AI 2021 Challenge: Report39
This paper introduces the first Mobile AI challenge, where the target is to develop an end-to-end deep learning-based depth estimation solutions that can demonstrate a nearly realtime performance on smartphones and IoT platforms.
- Fast Camera Image Denoising on Mobile GPUs with Deep Learning, Mobile AI 2021 Challenge: Report39
The first Mobile AI challenge is introduced, where the target is to develop an end-to-end deep learning-based image denoising solution that can demonstrate high efficiency on smartphone GPUs.
- Event-Based Eye Tracking. AIS 2024 Challenge Survey38
This survey reviews the AIS 2024 Event-Based Eye Tracking (EET) Challenge and examines novel and diverse methods from the submitted factsheets to advance future event-based eye tracking research.
- 38
- NTIRE 2025 Challenge on Day and Night Raindrop Removal for Dual-Focused Images: Methods and Results37
This paper reviews the NTIRE 2025 Challenge on Day and Night Raindrop Removal for Dual-Focused Images to establish a new and powerful benchmark for the task of removing raindrops under varying lighting and focus conditions.
- Artificial intelligence in medico-dental diagnostics of the face: a narrative review of opportunities and challenges37
This review confirms the many benefits derived from AI solutions used for the diagnosis of medical images, but highlights the evident lack of regulatory oversight, the urgent need to establish licensing protocols, and the imperative to investigate the moral quality of new norms set with the implementation of AI applications in medico-dental diagnostics.
- Basic Binary Convolution Unit for Binarized Image Restoration Network36
This study reconsider components in binary convolution, such as residual connection, BatchNorm, activation function, and structure, for IR tasks, and finds that residual connection can reduce the information loss caused by binarization, and BBCU can serve as a basic unit for binarized IR networks.
- NTIRE 2021 Challenge on Video Super-Resolution36
This paper presents evaluation results from two competition tracks as well as the proposed solutions to the NTIRE 2021 Challenge on Video Super-Resolution, and develops conventional video SR methods focusing on the restoration quality.
- NTIRE 2021 Depth Guided Image Relighting Challenge36
This paper reviews the NTIRE 2021 depth guided image relighting challenge and relies on the VIDIT dataset for each of the two challenge tracks, including depth information, to transform the illumination settings of an input image to match those of another guide image, similar to style transfer.
- 36
- 3D Appearance Super-Resolution With Deep Learning36
Experimental results demonstrate that the proposed networks successfully incorporate the 3D geometric information and super-resolve the texture maps.
- 36
- NTIRE 2022 Challenge on Stereo Image Super-Resolution: Methods and Results35
The 1st NTIRE challenge on stereo image super-resolution (restoration of rich details in a pair of low-resolution stereo images) with a focus on new solutions and results establishes a new benchmark for stereo image SR.
- DeepSEE: Deep Disentangled Semantic Explorative Extreme Super-Resolution35
This work proposes a novel framework, DeepSEE, for Deep disentangled Semantic Explorative Extreme super-resolution, and is the first method to leverage semantic maps for explorative super- resolution, and validates DeepSEE for up to 32x magnification and exploration of the space ofsuper-resolution.
- NTIRE 2020 Challenge on Image and Video Deblurring35
This challenge presents the evaluation results from 3 competition tracks as well as the proposed solutions, and the winning methods demonstrate the state-of-the-art performance on image and video deblurring tasks.
- The Tenth NTIRE 2025 Image Denoising Challenge Report34
An overview of the NTIRE 2025 Image Denoising Challenge is presented, highlighting the proposed methodologies and corresponding results, providing insights into the current state-of-the-art in image denoising.
- 34
- Single-Image super-resolution - When model adaptation matters34
This paper proposes a variation of deep residual convolutional neural networks, which has been carefully designed for robustness and efficiency in both learning and testing, and proposes multiple strategies for model adaptation to the internal contents of the low-resolution input image and analyzes their strong points and weaknesses.
- 34
- Learning Transformer-based World Models with Contrastive Predictive Coding33
It is shown that the next state prediction objective adopted in previous approaches is insufficient to fully exploit the representation capabilities of Transformers, and proposed to extend world model predictions to longer time horizons by introducing TWISTER (Transformer-based World model wIth contraSTivE Representations), a world model using action-conditioned Contrastive Predictive Coding to learn high-level temporal feature representations and improve the agent performance.
- 33
- 33
- Towards Efficient Graph Convolutional Networks for Point Cloud Handling33
A series of experiments show that the optimized networks have reduced computational complexity, decreased memory consumption, and accelerated inference speed while maintaining comparable accuracy for learning on point clouds.
- Semantic super-resolution: When and where is it useful?33
Experimental results show that the semantic driven super-resolution can significantly improve over the original settings, and the benefits vs. the drawbacks of using semantic information are discussed.
- COMBINING TRAFFIC SIGN DETECTION WITH 3D TRACKING TOWARDS BETTER DRIVER ASSISTANCE33
A realtime version of driver assistance systems that integrates single view detection with region-based 3D tracking of traffic signs and obtains 3D pose information that is used to establish the relevance of the traffic sign to the driver.
- 32
- 32
- 31
- Deep Equilibrium Diffusion Restoration with Parallel Sampling31
This work derives an analytical solution by modeling the entire sampling chain in these IR models as a joint multivariate fixed point system, called DeqIR, which can conduct parallel sampling and restore HQ images without training and compute fast gradients via DEQ inversion.
- 31
- 31
- 31
- NTIRE 2025 Challenge on Efficient Burst HDR and Restoration: Datasets, Methods, and Results30
A comprehensive overview of the NTIRE 2025 Efficient Burst HDR and Restoration Challenge is provided, the proposed solutions are compared, and it serves as a valuable reference for researchers and practitioners in efficient burst HDR and restoration.
- NTIRE 2025 Challenge on HR Depth From Images of Specular and Transparent Surfaces30
The NTIRE 2025 challenge on HR Depth From images of Specular and Transparent surfaces, held in conjunction with the New Trends in Image Restoration and Enhancement workshop at CVPR 2025, proposes two tracks on stereo and single-image depth estimation.
- 30
- 30
- PyNet-V2 Mobile: Efficient On-Device Photo Processing With Neural Networks30
The results demonstrate that the PyNET-V2 Mobile model can substantially surpass the quality of tradition ISP pipelines, while outperforming the previously introduced neural network-based solutions designed for fast image processing.
- Optuna vs Code Llama: Are LLMs a New Paradigm for Hyperparameter Tuning?29
This work explores the use of large language models (LLMs) for hyperparameter optimization by fine-tuning a parameter-efficient version of Code Llama using LoRA, demonstrating that LLM-based optimization not only rivals established Bayesian methods like Tree-structured Parzen Estimators (TPE), but also accelerates tuning for real-world applications requiring perceptual quality and low-latency processing.
- NTIRE 2025 Challenge on Low Light Image Enhancement: Methods and Results29
This paper thoroughly evaluates the state-of-the-art advancements in LLIE, showcasing the significant progress.
- 29
- 29
- 29
- 29
- From Face Images and Attributes to Attributes29
It is shown that handcrafted features can be as good as deep features, that the attributes themselves are powerful enough to predict other attributes and that clustering the samples according to their attributes can mitigate the training complexity for deep learning.
- 29
- NTIRE 2025 Challenge on Event-Based Image Deblurring: Methods and Results28
This paper presents an overview of NTIRE 2025, the First Challenge on Event-Based Image Deblurring, detailing the proposed methodologies and corresponding results, and anticipate that this challenge will drive further advancements in event-based vision research.
- NTIRE 2025 Challenge on Raw Image Restoration and Super-Resolution28
The NTIRE 2025 RAW Image Restoration and Super-Resolution Challenge is reviewed, highlighting the proposed solutions and results, and presents the current state-of-the-art in RAW Restoration.
- 28
- 28
- Fast and Accurate Quantized Camera Scene Detection on Smartphones, Mobile AI 2021 Challenge: Report28
The first Mobile AI challenge is introduced, where the target is to develop quantized deep learning-based camera scene classification solutions that can demonstrate a real-time performance on smartphones and IoT platforms.
- A Weakly Supervised Convolutional Network for Change Segmentation and Classification28
W-CDNet is presented, a novel weakly supervised change detection network that can be trained with image-level semantic labels and achieves top results on AICD and HRSCD, two public aerial imaging change detection datasets as well as on a Food Waste change detection dataset.
- AIM 2020 Challenge on Image Extreme Inpainting28
This paper reviews the AIM 2020 challenge on extreme image inpainting and proposes solutions and results for two different tracks: classical image inPainting and semantically guided image in Painting.
- Naive Bayes Image Classification: Beyond Nearest Neighbors28
In this paper, the replacement of the nearest neighbor part with more elaborate and robust (sparse) representations are studied, as well as trading performance for speed for practical purposes.
- 28
- The Eleventh NTIRE 2026 Efficient Super-Resolution Challenge Report27
This paper reviews the NTIRE 2026 challenge on efficient single-image super-resolution with a focus on the proposed solutions and results and gauges the state-of-the-art results for efficient single-image super-resolution.
- 27
- 27
- NTIRE 2021 Challenge on Quality Enhancement of Compressed Video: Dataset and Study27
A novel dataset for video enhancement and the state-of-the-art methods of the NTIRE 2021 challenge on quality enhancement of compressed video are studied.
- Robust Deepfake Detection, NTIRE 2026 Challenge: Report26
This report presents the competition setting, dataset preparation, as well as details and performance of methods, which rely on large foundation models, ensembles, and degradation training to combine generality and robustness in deepfake detection.
- Ntire 2025 Challenge on Short-Form Ugc Video Quality Assessment and Enhancement: Methods and Results26
This paper presents a review for the NTIRE 2025 Challenge on Short-form UGC Video Quality Assessment and Enhancement, and introduces a new short-form UGC dataset tailored for single-image super-resolution, i.e., the KwaiSR dataset.
- NTIRE 2025 Challenge on Real-World Face Restoration: Methods and Results26
This paper provides a review of the NTIRE 2025 challenge on real-world face restoration, highlighting the proposed solutions and the resulting outcomes and offers an in-depth overview of the latest trends in the field.
- 26
- 26
- The First Challenge on Remote Sensing Infrared Image Super-Resolution at NTIRE 2026: Benchmark Results and Method Overview25
The NTIRE 2026 Remote Sensing Infrared Image Super-Resolution (x4) Challenge, one of the associated challenges of NTIRE 2026, aims to recover high-resolution infrared images from low-resolution inputs generated through bicubic downsampling with a x4 scaling factor.
- 25
- 25
- 25
- 25
- 25
- Single Image Depth Prediction Made Better: A Multivariate Gaussian Take25
Per-pixel depth covariance modeling leads to a computationally expensive continuous loss function, which the SIDP model obtained by optimizing the learned low-rank approximation of the overall covariance matrix shows state-of-the-art results.
- 25
- NTIRE 2025 Challenge on UGC Video Enhancement: Methods and Results24
An overview of the NTIRE 2025 Challenge on UGC Video Enhancement is presented, which constructed a set of 150 user-generated content videos without reference ground truth, which suffer from real-world degradations such as noise, blur, faded colors, compression artifacts, etc.
- NTIRE 2025 the 2nd Restore Any Image Model (RAIM) in the Wild Challenge24
This challenge established a new benchmark for real-world image restoration, featuring diverse scenarios with and without reference ground truth, and tasked participants with restoring real-captured images suffering from complex and unknown degradations.
- Arbitrary-Scale Image Synthesis24
This work proposes the design of scale-consistent positional encodings invariant to the generator's layers transformations that enables the generation of arbitrary-scale images even at scales unseen during training.
- 3D-Aware Video Generation24
This work develops a GAN framework that synthesizes 3D video supervised only with monocular videos and learns a rich embedding of decomposable 3D structures and motions that enables new visual effects of spatio-temporal renderings while producing imagery with quality comparable to that of existing 3D or video GANs.
- AIM 2020 Challenge on Video Extreme Super-Resolution: Methods and Results24
This paper reviews the video extreme super-resolution challenge associated with the AIM 2020 workshop at ECCV 2020 and gauges the state-of-the-art for such a demanding task, where fidelity to the ground truth is measured by PSNR and SSIM.
- PIRM2018 Challenge on Spectral Image Super-Resolution: Methods and Results24
The Perceptual Image Restoration and Manipulation (PIRM) workshop challenge on spectral image super-resolution is described, the methods used by the participants, summarise the results and discuss their rankings.
- Raw Image Reconstruction From Rgb on Smartphones. Ntire 2025 Challenge Report23
This paper aims to recover RAW sensor images from smartphones given the corresponding sRGB images without metadata and, by doing this, “reverse” the ISP transformation.
- 23
- NTIRE 2025 Challenge on Image Super-Resolution ($\times 4$): Methods and Results23
The NTIRE 2025 image super-resolution ($\times 4$) challenge, one of the associated competitions of the 10th NTIRE Workshop at CVPR 2025, aims to recover high-resolution images from low-resolution counterparts generated through bicubic downsampling with a $\times 4$ scaling factor.
- LEMUR Neural Network Dataset: Towards Seamless AutoML23
LEMUR is an open-source dataset and framework that provides a large collection of PyTorch-based neural networks across tasks such as classification, segmentation, detection, and natural language processing and is extensible, allowing researchers to add new models, datasets, or metrics without breaking compatibility.
- NTIRE 2024 Challenge on Stereo Image Super-Resolution: Methods and Results23
This paper summarizes the 3rd NTIRE challenge on stereo image super-resolution with a focus on new solutions and results and establishes a new benchmark for stereo image SR.
- 23
- Using artificial intelligence to determine the influence of dental aesthetics on facial attractiveness in comparison to other facial modifications23
Alignment of teeth improves facial attractiveness to a similar extent than wearing lipstick, but has no discernable effect on perceived age.
- 23
- 23
- Ntire 2025 Image Shadow Removal Challenge Report22
This work examines the findings of the NTIRE 2025 Shadow Removal Challenge, with 17 teams successfully submitting their solutions during the final evaluation phase, and one focusing on reconstruction fidelity and the other on visual perception through a user study.
- What You Have is What You Track: Adaptive and Robust Multimodal Tracking22
This paper proposes a flexible framework for robust multimodal tracking through a novel Heterogeneous Mixture-of-Experts fusion mechanism with adaptive complexity, coupled with a video-level masking strategy that ensures both temporal consistency and spatial completeness - critical for effective video tracking.
- 22
- NTIRE 2024 Restore Any Image Model (RAIM) in the Wild Challenge22
The RAIM challenge constructed a benchmark for image restoration in the wild, including real-world images with/without reference ground truth in various scenarios from real applications, which improved the state-of-the-art restoration performance and obtained unanimous recognition from all 18 judges.
- 22
- AIM 2020 Challenge on Video Temporal Super-Resolution22
This paper reports the second AIM challenge on Video Temporal Super-Resolution (VTSR), a.k.a. frame interpolation, with a focus on the proposed solutions, results, and analysis, and proposes the enhanced quadratic video interpolation method.
- PIRM2018 Challenge on Spectral Image Super-Resolution: Dataset and Study22
This paper introduces a newly collected and novel dataset (StereoMSI) for example-based single and colour-guided spectral image super-resolution, which is the first of its kind, comprising 350 registered colour-spectral image pairs.
- 22
- 21
- 21
- 21
- Deep RAW Image Super-Resolution. A NTIRE 2024 Challenge Survey21
The NTIRE 2024 RAW Image Super-Resolution Challenge is reviewed, highlighting the proposed solutions and results, and providing a gauge for the current state-of-the-art in RAW Image Super-Resolution.
- MoVideo: Motion-Aware Video Generation with Diffusion Models21
A novel motion-aware video generation (MoVideo) framework that takes motion into consideration from two aspects: video depth and optical flow, showing promising prompt consistency, frame consistency and visual quality.
- 21
- The Heterogeneity Hypothesis: Finding Layer-Wise Differentiated Network Architectures21
This paper investigates a design space that is usually overlooked, i.e. adjusting the channel configurations of predefined networks, and finds that this adjustment can be achieved by shrinking widened baseline networks and leads to superior performance.
- 21
- Extreme Learned Image Compression with GANs21
A user study confirms that for low bitrates, this approach significantly outperforms state-of-the-art methods, saving up to 67% compared to the next-best method BPG.
- Iterative Nearest Neighbors21
A novel sparse representation, the Iterative Nearest Neighbors (INN), that combines the power of SR and LLE with the computational simplicity of k NN is proposed and proves on par or better performance with MP and OMP on sparse signal recovery task.
- 20
- Multilingual Audio-Visual Speech Recognition with Hybrid CTC/RNN-T Fast Conformer20
This work adapts the recently proposed Fast Conformer model to process both audio and visual modalities using a novel hybrid CTC/RNN-T architecture and demonstrates the ability of the proposed model to perform audio-only, visual-only, and audio-visual speech recognition at test time.
- Deep Portrait Quality Assessment. A NTIRE 2024 Challenge Survey20
The NTIRE 2024 Portrait Quality Assessment Challenge is reviewed, highlighting the proposed solutions and results, and providing a gauge for the current state-of-the-art in Portrait Quality Assessment.
- 20
- 20
- 20
- 20
- Fast Perceptual Image Enhancement20
This work builds upon the prior work and explores different network architectures targeting an increase in image quality and speed and achieves a significantly higher mean opinion score (MOS) than the baseline while speeding up the computation by 6.3 times on a consumer-grade CPU.
- AIM 2019 Challenge on Video Temporal Super-Resolution: Methods and Results20
This paper reviews the first AIM challenge on video temporal super-resolution (frame interpolation) with a focus on the proposed solutions and results, and employs the REDS_VTSR dataset derived from diverse videos captured in a hand-held camera for training and evaluation purposes.
- 19
- NTIRE 2026 Challenge on Bitstream-Corrupted Video Restoration: Methods and Results19
The NTIRE 2026 Challenge on Bitstream-Corrupted Video Restoration provides a common benchmark for evaluating restoration methods under realistic corruption settings and provides useful insights for future research on robust video restoration under practical bitstream corruption.
- Generative Flows with Invertible Attentions19
Two types of invertible attention mechanisms are introduced, i.e., map-based and transformer-based attentions, for both unconditional and conditional generative flows, to exploit a masked scheme of these two attentions to learn long-range data dependencies in the context ofGenerative flows.
- Generating Masks from Boxes by Mining Spatio-Temporal Consistencies in Videos19
This work introduces a method for generating segmentation masks from per-frame bounding box annotations in videos, and proposes a spatio-temporal aggregation module that effectively mines consistencies in the object and background appearance across multiple frames.
- Learning to Improve Image Compression Without Changing the Standard Decoder19
A frequency-domain pre-editing method is proposed to optimize the distribution of DCT coefficients, aiming at facilitating the JPEG compression, and it is proposed that this translates to better overall color retention especially when strong compression is applied.
- 18
- 17
- Conformer and Blind Noisy Students for Improved Image Quality Assessment17
This work explores the performance of transformer-based full- reference IQA models and proposes a method for IQA based on semi-supervised knowledge distillation from full-reference teacher models into blind student models using noisy pseudo-labeled data.
- Generalized Real-World Super-Resolution through Adversarial Robustness17
This work presents Robust Super-Resolution (RSR), a method that leverages the generalization capability of adversarial attacks to tackle real-world SR, and demonstrates that the RSR method generalizes well across datasets without re-training for specific noise priors.
- 17
- 17
- AugmentGest: Can Random Data Cropping Augmentation Boost Gesture Recognition Performance?16
A comprehensive data augmentation framework that integrates geometric transformations, random cropping, rotation, zooming and intensity-based transformations, brightness and contrast adjustments to simulate real-world variations and offers a scalable solution to advance HGR and action recognition applications in real-world scenarios is proposed.
- ReCap: Better Gaussian Relighting with Cross-Environment Captures16
This work presents ReCap, treating cross-environment captures as multi-task target to provide the missing supervision that cuts through the entanglement, and jointly optimizes multiple lighting representations that share a common set of material attributes.
- 16
- VIST-GPT: Ushering in the Era of Visual Storytelling with LLMs?16
This paper utilizes RoViST and GROOVIST, novel reference-free metrics designed to assess visual storytelling, focusing on visual grounding, coherence, and non-redundancy, to provide a more nuanced evaluation of narrative quality.
- 16
- 16
- Normalizing Flow as a Flexible Fidelity Objective for Photo-Realistic Super-resolution16
This work revisiting the L1 loss and showing that it corresponds to a one-layer conditional flow is shown to outperform state-of-the-art methods for photo-realistic super-resolution and demonstrate that the flexibility of deeper flows leads to better visual quality and consistency when combined with adversarial losses.
- NTIRE 2020 Challenge on Video Quality Mapping: Methods and Results16
This paper reviews the NTIRE 2020 challenge on video quality mapping (VQM), which addresses the issues of quality mapping from source video domain to target video domain, and some existing methods are evaluated, showing promising solutions to the weakly-supervised videoquality mapping problem.
- Adversarial feature distribution alignment for semi-supervised learning16
This paper shows that a consequence of overfitting in SSL is feature distribution misalignment between labeled and unlabeled samples, and proposes a new feature distribution alignment method that is particularly effective when using only a small amount of labeled samples.
- 16
- 16
- 16
- Stixels Motion Estimation without Optical Flow Computation16
This paper presents a new approach to estimate the motion of objects seen from a stereo rig mounted on a ground mobile robot, and exploits the prior knowledge on ground plane presence and rough shape of objects, to extract a simplified world model, named stixel world.
- Bokehlicious: Photorealistic Bokeh Rendering with Controllable Apertures15
This work proposes Bokehlicious, a highly efficient network that provides intuitive control over Bokeh strength through an ApertureAware Attention mechanism, mimicking the physical lens aperture, achieving competitive results on the RealDOF benchmark.
- NTIRE 2024 Challenge on Night Photography Rendering15
This paper presents a review of the NTIRE 2024 challenge to find solutions that process raw camera images taken in nighttime conditions, and thereby produce a photo-quality output images in the standard RGB (sRGB) space.
- 15
- 15
- 15
- 15
- From Memorization to Creativity: LLM as a Designer of Novel Neural Architectures14
It is demonstrated that iterative self-supervised fine-tuning reshapes an LLM into a task-specialized architectural prior -- improving generation reliability, proxy performance, and structural diversity -- offering a reproducible, annotation-free alternative to hand-crafted search spaces.
- 14
- Versatile Vision Foundation Model for Image and Video Colorization14
This work shows how a latent diffusion model, pre-trained on text-to-image synthesis, can be finetuned for image colorization and provide a flexible solution for a wide variety of scenarios: high quality direct colorization with diverse results, user guided colorization through colors hints, text prompts or reference image and finally video colorization.
- 14
- Tracking the Known and the Unknown by Leveraging Semantic Information.14
This work proposes a tracking framework that can exploit semantic information, without sacrificing the generic nature of the tracker, and model the class of the object through a semantic module that provides complementary class-specific predictions.
- 14
- An Elastic Deformation Field Model for Object Detection and Tracking14
Experiments show that the deformation field can better approximate real object deformations and therefore, for certain classes, produces even better detection accuracy than state-of-the-art DPM.
- Resource-Efficient Iterative LLM-Based NAS with Feedback Memory13
This work proposes a closed-loop pipeline leveraging large language models (LLMs) to iteratively generate, evaluate, and refine convolutional neural network architectures for image classification on a single consumer-grade GPU without LLM fine-tuning, establishing a low-budget, reproducible, and hardware-aware paradigm for LLM-driven NAS without cloud infrastructure.
- A Retrieval-Augmented Generation Approach to Extracting Algorithmic Logic from Neural Networks13
N-RAG is introduced, a retrieval-augmented generation system that converts large, heterogeneous PyTorch codebases into a searchable and executable library of validated neural modules, offering a first open-source solution that both quantifies and expands the diversity of executable neural architectures across repositories.
- Event-Free Moving Object Segmentation from Moving Ego Vehicle13
This work introduces a novel large-scale dataset called DSEC-MOS for moving object segmentation from moving ego vehicles, which is the first of its kind, and devise EmoFormer, a novel network able to exploit the event data, which relies only on event data for training but does not require event input during inference.
- AIS 2024 Challenge on Video Quality Assessment of User-Generated Content: Methods and Results13
This paper reviews the AIS 2024 Video Quality Assessment (VQA) Challenge, focused on User-Generated Content (UGC), to gather deep learning-based methods capable of estimating the perceptual quality of UGC videos.
- BSRAW: Improving Blind RAW Image Super-Resolution13
This work designs a realistic degradation pipeline tailored specifically for training models with raw sensor data that can upscale real-scene RAW images and improve their quality, and presents a new DSLM dataset and benchmark for this task.
- MuDreamer: Learning Predictive World Models without Reconstruction13
MuDreamer is presented, a robust reinforcement learning agent that builds upon the DreamerV3 algorithm by learning a predictive world model without the need for reconstructing input signals, and finds that the use of batch normalization is crucial to prevent learning collapse.
- Efficient and robust eye images iris segmentation using a lightweight U-net convolutional network13
The paper presents an efficient lightweight U-net convolutional neural network (CNN) architecture that can be used for iris segmentation in eye images that leads to a much lower training time and eye image segmentation time (< 1 ms per image on Xeon CPU).
- 13
- 13
- 13
- SF‐YOLO: A Novel YOLO Framework for Small Object Detection in Aerial Scenes12
A novel YOLO framework called SF‐YOLO for small object detection is proposed, which performs weighted fusion on feature maps by combining fast normalized fusion method and CARAFE operation, accurately assessing the importance of each feature and enhancing the representation of small objects.
- 12
- Efficient Degradation-aware Any Image Restoration12
Extensive comparisons demonstrate that the DaAIR outperforms both state-of-the-art All-in-One models and degradation-specific counterparts, affirming the efficacy and practicality of this efficient restorer integrating holistic and specific learning within a unified model.
- 12
- Unsupervised Multimodal Video-to-Video Translation via Self-Supervised Learning12
UVIT, a novel unsupervised video-to-video translation model that decomposes the style and the content, uses the specialized encoder-decoder structure and propagates the inter-frame information through bidirectional recurrent neural network (RNN) units, enables to achieve style consistent video translation results.
- Efficiently Detecting Plausible Locations for Object Placement Using Masked Convolutions12
This paper proposes a fast method for the automatic selection of plausible locations for object insertion into images using masked convolutions to compute featuremaps for left, right, top and bottom contexts just once per image.
- NTIRE 2026 3D Restoration and Reconstruction in Real-world Adverse Conditions: RealX3D Challenge Results11
This paper presents a comprehensive review of the NTIRE 2026 3D Restoration and Reconstruction (3DRR) Challenge, detailing the proposed methods and results and thoroughly evaluate the submitted approaches against state-of-the-art baselines, revealing significant progress in 3D reconstruction under adverse conditions.
- 11
- 11
- 11
- 11
- Fast and Accurate Camera Scene Detection on Smartphones11
A novel Camera Scene Detection Dataset (CamSDD) containing more than 11K manually crawled images belonging to 30 different scene categories is proposed and an efficient and NPU-friendly CNN model is proposed that demonstrates a top-3 accuracy of 99.5% on this dataset and achieves more than 200 FPS on the recent mobile SoCs.
- Extremely Weak Supervised Image-to-Image Translation for Semantic Segmentation11
The experiments show that an extremely weak supervised I2I translation solution using only one paired training sample can achieve a quantitative performance much better than the unsupervised CycleGAN model, and comparable to that of the supervised pix2pix model trained on thousands of pairs.
- 11
- MatIR: A Hybrid Mamba-Transformer Image Restoration Model10
A Mamba-Transformer hybrid image restoration model called MatIR is proposed, which cross-cycles the blocks of the Transformer layer and the Mamba layer to extract features, thereby taking full advantage of the advantages of the two architectures.
- 10
- 10
- 10
- An Efficient Recurrent Adversarial Framework for Unsupervised Real-Time Video Enhancement10
This work proposes an efficient adversarial video enhancement framework that learns directly from unpaired video examples and introduces new recurrent cells that consist of interleaved local and global modules for implicit integration of spatial and temporal information.
- 10
- Deep Line Encoding for Monocular 3D Object Detection and Depth Prediction10
The deep line encoding is proposed to make better use of the line information in scenes and advances the state-of-the-art on KITTI single-image 3D object detection and depth prediction benchmarks.
- 10
- Failure Detection for Facial Landmark Detectors10
This work study two top recent facial landmark detectors and devise confidence models for their outputs and validate their failure detection approaches on standard benchmarks and correctly identify more than 40% of the failures in the outputs of the landmark detectors.
- PICASO: PIxel correspondences and SOft match selection for real-time tracking10
PICASO, an efficient, yet strongly performing tracking scheme that evaluates its tracker on standard videos for rigid and non-rigid object tracking, and obtains excellent performance at 42fps with Matlab on a CPU.
- Automatic Stave Discovery for Musical Facsimiles10
The staves are represented by repetitive line patterns, are more constrained and informative, and thus this work proposes direct optimization over such patterns instead of first spotting single staff lines, and the optimal binarization threshold also is the one giving the maximum evidence for the presence of staves.
- 9
- 9
- Parameterized Low-Rank Regularizer for High-dimensional Visual Data9
A Parameterized Low-Rank Regularizer (PLRR) is proposed, which models low-rank visual data through matrix factorization by utilizing neural networks to parameterize the factor matrices, whose feasible domains are essentially constrained.
- AIM 2025 Rip Current Segmentation (RipSeg) Challenge Report9
This report presents an overview of the AIM 2025 RipSeg Challenge, a competition designed to advance techniques for automatic rip current segmentation in still images, and outlines the dataset details, competition framework, evaluation metrics, and final results.
- 9
- Towards closing the gap in weakly supervised semantic segmentation with DCNNs: Combining local and global models9
A systematic study of the different ingredients involved in the weakly supervised scenario and an original experimental strategy unravel a counter-intuitive mechanism that is simple and amenable to generalisations to other weakly-supervised scenarios: averaging poor local predicted annotations with the baseline ones and reuse them for training a DCNN yields new state-of-the-art results.
- 9
- 9
- 9
- 9
- 8
- 8
- Virtually Enriched NYU Depth V2 Dataset for Monocular Depth Estimation: Do We Need Artificial Augmentation?8
This is the first work that augments a real-world dataset with randomly generated virtual 3D objects for monocular depth estimation and it is shown that ANYU improves the monocular depth estimation performance and generalization of deep neural networks with considerably different architectures.
- Toward Efficient Deep Blind Raw Image Restoration8
A new realistic degradation pipeline for training deep blind RAW restoration models that considers realistic sensor noise, motion blur, camera shake, and other common degradations and is considered to be the most exhaustive analysis on RAW image restoration.
- Towards Online Real-Time Memory-based Video Inpainting Transformers8
This work proposes a framework to adapt existing inpainting transformers to these constraints by memorizing and refining redundant computations while maintaining a decent inpainting quality and shows great online results with a consistent throughput above 20 frames per second.
- 8
- Deep Learning for Visual Data Compression8
The recent progress in deep learning based visual data compression, including image compression, video compression and point cloud compression is introduced.
- Self-Supervised 2D Image to 3D Shape Translation with Disentangled Representations8
SIST is proposed, a Self-supervised Image to Shape Translation framework that fulfills three tasks: reconstructing the 3D shape from a single image; learning disentangled representations for shape, appearance and viewpoint; and generating a realistic RGB image from these independent factors.
- Learned Lightweight Smartphone ISP with Unpaired Data7
This work proposes a novel training method for a learnable ISP that eliminates the need for direct correspondences between raw images and ground-truth data with matching content and employs a multi-term loss function guided by adversarial training with multiple discriminators processing feature maps from pre-trained networks to maintain content structure.
- Accurate and Efficient World Modeling with Masked Latent Transformers7
EMERALD (Efficient MaskEd latent tRAnsformer worLD model), a world model using a spatial latent state with MaskGIT predictions to generate accurate trajectories in latent space and improve the agent performance is introduced.
- 7
- A Study of Forward-Forward Algorithm for Self-Supervised Learning7
The main finding is that while the forward-forward algorithm performs comparably to backpropagation during (self-)supervised training, the transfer performance is significantly lagging behind in all the studied settings.
- Editorial: Introduction to the Issue on Deep Learning for Image/Video Restoration and Compression7
This special issue covers the state of the art in learned image/video restoration and compression to promote further progress in innovative architectures and training methods for effective and efficient networks for image/video restoration and compression.
- SMILE: Semantically-guided Multi-attribute Image and Layout Editing7
This paper successfully exploit a multimodal representation that handles all attributes, be it guided by random noise or exemplar images, while only using the underlying domain information of the target domain.
- Fast Few-Shot Classification by Few-Iteration Meta-Learning7
This work is the first to integrate both induction and transduction into the base learner in an optimization-based meta-learning framework, and demonstrates the speed and effectiveness of this approach on four few-shot classification datasets.
- 7
- 7
- Flexible Example-Based Image Enhancement with Task Adaptive Global Feature Self-guided Network7
This model outperforms the current state of the art in learning a single enhancement mapping, while having significantly fewer parameters than its competitors, and achieves even higher performance on learning multiple mappings simultaneously, by taking advantage of shared representations.
- 7
- $$k^2$$-means for Fast and Accurate Large Scale Clustering7
In its extensive experiments, the proposed divisive initialization generally leads to clustering energies comparable to those achieved with the standard k-means++ initialization, while being significantly faster.
- Illuminating Darkness: Learning to Enhance Low-light Images In-the-Wild6
The Low-Light Smartphone Dataset (LSD), a large-scale, high-resolution (4K+) dataset collected in the wild across a wide range of challenging lighting conditions, is introduced and TFFormer, a hybrid model that encodes luminance and chrominance separately to reduce color-structure entanglement is proposed.
- Color Matching Using Hypernetwork-Based Kolmogorov-Arnold Networks6
A hypernetwork is developed that generates spatially varying weight maps to control the nonlinear splines of a KAN, enabling accurate color matching and outperforms existing approaches while remaining lightweight compared to other methods.
- 6
- The Heterogeneity Hypothesis: Finding Layer-Wise Dissimilated Network Architecture.6
The ``heterogeneity hypothesis'' is articulate: with the same training protocol, there exists a layer-wise dissimilated network architecture (LW-DNA) that can outperform the original network with regular channel configurations under lower level of model complexity.
- 6
- 6
- Generic 3D Convolutional Fusion for Image Restoration6
The proposed 3DCF method uses the exact same convolutional network architecture to address both image denoising and single image super-resolution, and achieves substantial improvements over the state of the art methods that it fuses on standard benchmarks for both tasks.
- 5
- ContextFormer: Redefining Efficiency in Semantic Segmentation5
This work proposes ContextFormer, a hybrid framework leveraging the strengths of CNNs and ViTs in the bottleneck to balance efficiency, accuracy, and robustness for real-time semantic segmentation, setting a new benchmark for efficiency and performance.
- 5
- 5
- Scoring facial attractiveness with deep convolutional neural networks: How training on standardized images reduces the bias of facial expressions5
Deep convolutional neural networks are applied to identify how facial expressions affect facial attractiveness and to explore whether a dedicated training of the CNN is able to reduce the bias of facial expressions.
- 5
- 5
- Leveraging observation uncertainty for robust visual tracking5
The accuracy of visual tracking is enhanced by leveraging a novel measure for observation quality by preprocessing the image by subtracting the background, detecting salient regions, and rendering the image illumination-invariant.
- INRetouch: Context Aware Implicit Neural Representation for Photography Retouching4
A novel retouch transfer approach that learns from professional edits through before-after image pairs, enabling precise replication of complex editing operations and a context-aware Implicit Neural Representation that learns to apply edits adaptively based on image content and context.
- 4
- CH-YOLO-Lite: a lightweight object detection model with context-aware progressive aggregation and hierarchical feature aggregation for aerial imagery4
This work designed a context-aware progressive aggregator (CPA) to integrate multi-scale contextual information through a progressive fusion strategy, enhancing the model’s perception of its environment and improving small object detection in complex scenes and significantly reduces the parameters of the model while maintaining the detection performance.
- MIORe & VAR-MIORe: Benchmarks to Push the Boundaries of Restoration4
MIORe and VAR-MIORe are introduced, two novel multi-task datasets that address critical limitations in current motion restoration benchmarks and provide high-resolution, scalable ground truths that challenge existing algorithms under both controlled and adverse conditions.
- 4
- RipVIS: Rip Currents Video Instance Segmentation Benchmark for Beach Monitoring and Safety4
RipVIS aims to set a new standard for rip current segmentation, contributing towards safer beach environments and offers a benchmark website to share data, models, and results with the research community, encouraging ongoing collaboration and future contributions.
- After the Party: Navigating the Mapping from Color to Ambient Lighting4
CL3AN is introduced, the first large-scale, high-resolution dataset of its kind designed to facilitate the restoration of images captured under multiple Colored Light sources to their AmbientNormalized counterparts, and demonstrates enhanced robustness under non-homogeneous color lighting and material-specific reflectance variations, all while maintaining a highly competitive computational cost.
- 4
- Restore Anything Model via Efficient Degradation Adaptation4
A unified manner to achieve joint embedding is proposed by leveraging the inherent similarities across various degradations for efficient and comprehensive restoration by leveraging the inherent similarities across various degradations.
- 4
- PASTA: Towards Flexible and Efficient HDR Imaging Via Progressively Aggregated Spatio-Temporal Alignment4
PASTA, a novel Progressively Aggregated Spatio-Temporal Alignment framework for HDR deghosting achieves effectiveness and efficiency by harnessing hierarchical representation during feature distanglement and substantially boosts computational speed and optimizes the HDR imaging workflow.
- 4
- Photo-Realistic and Robust Inpainting of Faces Using Refinement GANs4
This paper proposes, design and evaluates a refinement generative adversarial network (GAN), that is applied within an inpainting pipeline to robustly generate photo-realistic restorations and concludes, that refinement considerably improves for each setting the numerical and visual quality and preserves already good restoration.
- 4
- 4
- Restore, Assess, Repeat: A Unified Framework for Iterative Image Restoration3
This work proposes RAR, a Restore, Assess and Repeat process, that integrates Image Quality Assessment (IQA) and Image Restoration (IR) into a unified framework to iteratively and efficiently achieve high quality image restoration.
- AdaptSR: Low-Rank Adaptation for Efficient and Scalable Real-World Super-Resolution3
This work introduces AdaptSR, a low-rank adaptation (LoRA) framework that efficiently repurposes bicubic-trained SR models for real-world tasks and matches or exceeds full model fine-tuning while training 92% fewer parameters, enabling rapid adaptation to real SR tasks within minutes.
- Higher Fidelity Perceptual Image and Video Compression with a Latent Conditioned Residual Denoising Diffusion Model3
A hybrid compression scheme optimized for perceptual quality is proposed, extending the approach of the CDC model with a decoder network in order to reduce the impact on distortion metrics such as PSNR.
- 3
- DiTVR: Zero-Shot Diffusion Transformer for Video Restoration3
DiTVR is introduced, a zero shot video restoration framework that couples a diffusion transformer with trajectory aware attention and a wavelet guided, flow consistent sampler, demonstrating superior temporal consistency and detail preservation while remaining robust to flow noise and occlusions.
- Steering Prediction via a Multi-Sensor System for Autonomous Racing3
This work explores the integration of an event camera with the existing system to provide enhanced temporal information and introduces a new fusion learning policy to guide the fusion process, enhancing robustness against misalignment.
- 3
- Multi-Distillation Underwater Image Super-Resolution via Wavelet Transform3
A novel multi-distillation algorithm based on the swin transformer and wavelet transform and KDSWT lightweight (student) that significantly improves peak signal-to-noise ratio (PSNR) over the ( original) teacher model while decreasing the number of parameters and the calculation cost by 22%.
- 3
- 3
- Four Color Theorem for Fast Early Vision3
This work shows how to reduce the computational complexity of belief propagation by applying the Four Color Theorem to limit the maximum number of labels in the underlying image segmentation to at most four.
- Layover or Direct Flight: Rethinking Audio-Guided Image Segmentation2
This work focuses on object grounding, i.e., localizing an object of interest in a visual scene based on verbal human instructions based on verbal human instructions, and demonstrates that direct grounding from audio is not only feasible but, in some cases, even outperforms transcription-based methods, especially in terms of robustness to linguistic variability.
- CTSR: Controllable Fidelity-Realness Trade-off Distillation for Real-World Image Super Resolution2
Experiments conducted on several real-world image super-resolution benchmarks demonstrate that the proposed distillation-based approach surpasses existing state-of-the-art approaches, achieving superior performance across both fidelity and realness metrics.
- Improving Vision-Language Cross-Lingual Transfer with Scheduled Unfreezing2
Experiments with two mVLMs on three downstream tasks show that SUF brings consistent gains in ZS-XLT, especially for visual Q&A (xGQA) by up to 10 points, and the fact that in multilingual models these representations encode higher-level semantics rather than low-level language-specific idiosyncrasies should render SUF beneficial for ZS-XLT.
- Efficient End-to-End Convolutional Architecture for Point-of-Gaze Estimation2
This work introduces a methodology for data acquisition that shows promise due to its nonrestrictive and straightforward nature, notably increasing the yield of collected data without compromising diversity or quality, and presents a novel and efficient convolutional neural network specifically tailored for calibration-free point-of-gaze estimation.
- SOTA: Sequential Optimal Transport Approximation for Visual Tracking in Wild Scenario2
A probabilistic visual tracking method tailored for wild scenarios, where tracking environments experience abrupt changes over time, that considerably outperforms other state-of-the-art methods across several benchmark dataset.
- MS-RANAS: Multi-Scale Resource-Aware Neural Architecture Search2
This paper proposes Multi-Scale Resource-Aware Neural Architecture Search (MS-RANAS), a one-shot architecture search approach, capable of offering fast and accurate predictions and that could be deployed on a low-memory, low-power system-on-chip.
- 2
- 2
- 2
- 2
- Efficient Loopy Belief Propagation Using the Four Color Theorem2
It is shown how to reduce the computational complexity of belief propagation by applying the Four Color Theorem to limit the maximum number of labels in the underlying image segmentation to at most four.
- Robust Scene Stitching in Large Scale Mobile Mapping2
A novel loop-closure technique that is able to detect matches in challenging wide baseline conditions, were other systems tend to fail, and demonstrate robustness to odometric inconsistencies resulting from poor prior model build-up is presented.
- Multiple Scale Latents for Learned Image Compression1
This work proposes a hierarchical latent representation to improve the efficiency of the entropy model by using multiple latents at different scales, each with its own entropy model, to better capture the spatial structure of the latent representation.
- Mixture-of-Experts-Based Entropy Model for Learned Image Compression1
This paper proposes a MoE-based Entropy model (MoEE) for learned image compression, allowing the model to selectively activate only the subset of parameters required for the input image.
- Applying artificial intelligence to assess the impact of orthognathic treatment on gender-affirming facial features1
The results suggest that orthognathic corrections can significantly alter gender-affirming facial traits, which must be considered when defining individual treatment goals.
- Systematic Exploration of 4-Expert Heterogeneous Mixture-of-Experts via Automated Pipeline Search1
An automated large-scale search pipeline for heterogeneous 4-Expert Mixture-of-Experts (MoE4) architectures within the LEMUR neural network dataset ecosystem is presented and a critical finding emerged: due to alphabetical enumeration via itertools, the entire explored search space is anchored to a single family, AirNet.
- Winner of CVPR2026 NTIRE Challenge on Image Shadow Removal: Semantic and Geometric Guidance for Shadow Removal via Cascaded Refinement1
A three-stage progressive shadow-removal pipeline for the CVPR2026 NTIRE WSRD+ challenge, built on OmniSR, that combines RGB appearance with frozen DINOv2 semantic guidance and geometric cues from monocular depth and surface normals, reused across all stages.
- Convergence Theory for Iterative LLM-Based Neural Architecture Search: A Parametric Cross-Entropy Framework with Closed-Form Proxy Reliability1
Testing against a 22-cycle, three-LLM, six-dataset experiment with 3,300 generated architectures confirms two predictions quantitatively, two at direction-of-effect level, and explains the proxy-reliability ceiling effect previously reported empirically but left unexplained.
- 1
- 1
- 1
- 1
- A Simple Combination of Diffusion Models for Better Quality Trade-Offs in Image Denoising1
The proposed Linear Combination Diffusion Denoiser (LCDD), which unifies two complementary inference procedures-one that leverages the model's generative potential and another that ensures faithful signal recovery that ensures faithful signal recovery for additive Gaussian noise removal.
- 1
- Streaming Neural Images1
This work explores the critical yet overlooked limiting factors of INRs, such as computational cost, unstable performance, and robustness, and provides a deeper and more nuanced understanding of implicit neural image compression methods such as Fourier Feature Networks and Siren.
- 1
- Speeding Up Online Self-Supervised Learning by Exploiting Its Limitations1
This work proposes a framework for Online Self-supervised Learning with the goal of learning as much as possible from the newly arrived batch of data in a limited amount of training iterations before the next batch becomes available.
- 1
- 1
- 1
- Similarity-Guided Curriculum Fine-Tuning of LLMs for Neural Architecture Synthesis–
A cross-dataset transfer observation on SVHN, where direct base-model generation without curriculum warmup yields 27% peak SR at substantially lower accuracy than the CIFAR-10 equivalent, is reported, consistent with the increased synthesis difficulty of the unq-family anchor architecture.
- NTIRE 2026 Challenge on Efficient Low Light Image Enhancement: Methods and Results–
A systematic evaluation of recent methods for E-LLIE is provided, offering a comprehensive overview of state-of-the-art progress and demonstrating significant improvements in both performance and efficiency.
- –
- –
- –
- Scaling Closed-Loop Feature Channel Configuration with LLMs–
Findings indicate that the channel-search signal observed in the initial study transfers to a denser sampling regime and exposes architectural regularities that were difficult to assess from sparse observations.
- Geometry- and Physics-Aware Dataset Creation for Shadow Removal in High-Resolution Satellite Imagery–
- Curating Same-Family Neural Networks for LLM-Guided Model Improvement: A Controlled Case Study–
A source-guided candidate-generation protocol with non-source controls, source-conditioned candidates, and a no-LLM hp copy ablation under equal evaluation budgets is proposed for improving a weak target model using a stronger same-family source model from a neural-network database.
- –
- –
- –
- –
- Systematic Evaluation of Learning Rate Scheduling Strategies Across Heterogeneous Architectures–
The results show that the choice of scheduler depends heavily on the architecture: CosineAnnealingWarmRestarts and CyclicLR consistently outperform basic decay strategies.
- CLIP-Guided Multi-Task Regression for Multi-View Plant Phenotyping–
A level-aware vision language framework that jointly predicts plant age and leaf count using a single multi-task model built on CLIP embeddings that simplifies the pipeline by replacing the conventional dual-model setup while improving robustness to missing views.
- The RealDefocus Benchmark for Defocus Deblurring–
RealDefocus, a benchmark derived from the real-world RealBokeh dataset originally proposed for Bokeh Rendering, is built on, providing paired defocused inputs and sharp ground truth images, predefined training/validation/test splits, and a unified evaluation framework for comparing image restoration and neural rendering approaches.
- When Simplicity Wins: Bottleneck-Aware Context Modeling for Lightweight Semantic Segmentation–
Extensive experiments demonstrate that SiConMo achieves a state-of-the-art accuracy-efficiency trade-off among lightweight semantic segmentation models, highlighting simplicity as a powerful design principle.
- –
- –
- –
- –
- –
- DASR+: Training Domain Distance Aware Network for Unsupervised Image Super-Resolution–
A novel domain distance aware super-resolution (DASR+) approach with newly designed domain gap aware training strategy, domain distance weighted supervision strategy, and a domain distance adaptive network for unsupervised image super-resolution that outperforms state-of-the-art unsupervised SR methods in generating results with more realistic textures.
- TriFusion-SR: Joint Tri-Modal Medical Image Fusion and SR–
TriFusionSR is proposed, a wavelet-guided conditional diffusion framework for joint tri-modal fusion and super-resolution, enabling frequency-aware crossmodal interaction and structure-driven multimodal refinement.
- –
- Towards Robust Training in NNGPT AutoML Pipeline: A Loss-Optimizer Pairing Selection Study–
It is confirmed that no single pairing is universally optimal, and Cross-Entropy with Adam or AdamW is the most robust choice across architecture families and datasets, and actionable guidance for loss-optimizer selection within NNGPT Framework.
- –
- –
- –
- –
- –
- –
- Understanding the Performance of Deep Computer Vision Models: A Symbolic Regression Approach to Accuracy and Latency Prediction–
A two-stage statistical framework providing interpretable, closed-form insights into both architecture and training hyperparameters jointly determine accuracy and inference latency, and offers interpretable surrogate models for screening deep vision architectures under accuracy and latency constraints in edge deployment.
- –
- –
- –
- –
- LAFR: Efficient Diffusion-based Blind Face Restoration via Latent Codebook Alignment Adapter–
LAFR is a data and compute efficient one-step face-restoration method that improves perceptual quality, FID, landmark accuracy, and identity/structure consistency over OSEDiff-style baselines, without claiming uniform state of the art; it reaches this operating point while remaining competitive with far more expensive pipelines.
- –
- –
- –
- –
- –
- –
- –
- PIV3CAMS: a multi-camera dataset for multiple computer vision problems and its application to novel view-point synthesis–
This thesis introduces Paired Image and Video data from three CAMeraS, namely PIV3CAMS, aimed at multiple computer vision tasks and applies a model to the introduced PIV3CAMS dataset to synthesize novel target views as an example application of PIV3CAMS.
- –
- Zero-Pair Image to Image Translation using Domain Conditional Normalization–
This paper employs a single generator which has an encoder-decoder structure and analyzes different implementations of domain conditional normalization to obtain the desired target domain output to improve in qualitative and quantitative terms over the compared methods, while using much fewer parameters.
- Special Section Guest Editorial: Perceptually Optimized Imaging–
This guest editorial summarizes the Special Section on Perceptually Optimized Imaging and describes the current state of the art in perception-based image-guided medicine.
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.