Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks496 from public data
- Medical Transformer: Gated Axial-Attention for Medical Image Segmentation1,495
A Gated Axial-Attention model is proposed which extends the existing architectures by introducing an additional control mechanism in the self-attention module and achieves better performance than the convolutional and other related transformer-based architectures.
- HyperFace: A Deep Multi-Task Learning Framework for Face Detection, Landmark Localization, Pose Estimation, and Gender Recognition1,314
The proposed method, HyperFace, fuses the intermediate layers of a deep CNN using a separate CNN followed by a multi-task learning algorithm that operates on the fused features to exploit the synergy among the tasks which boosts up their individual performances.
- Densely Connected Pyramid Dehazing Network1,223
A new end-to-end single image dehazing method, called Densely Connected Pyramid Dehazing Network (DCPDN), which can jointly learn the transmission map, atmospheric light and dehazed all together by directly embedding the atmospheric scattering model into the network.
- Image De-Raining Using a Conditional Generative Adversarial Network1,218
This work attempts to leverage powerful generative modeling capabilities of the recently introduced conditional generative adversarial networks (CGAN) by enforcing an additional constraint that the de-rained image must be indistinguishable from its corresponding ground truth clean image.
- A Transformer-Based Siamese Network for Change Detection1,174
Experiments show that the proposed end-to-end trainable ChangeFormer architecture achieves better CD performance than previous counterparts.
- Density-Aware Single Image De-raining Using a Multi-stream Dense Network1,080
A novel density-aware multi-stream densely connected convolutional neural network-based algorithm, called DID-MDN, for joint rain density estimation and de-raining, which achieves significant improvements over the recent state-of-the-art methods.
- 899
- 876
- 872
- Generating High-Quality Crowd Density Maps Using Contextual Pyramid CNNs700
A novel method called Contextual Pyramid CNN (CP-CNN) for generating high-quality crowd density and count estimation by explicitly incorporating global and local contextual information of crowd images is presented.
- A survey of recent advances in CNN-based single image crowd counting and density estimation634
A comprehensive survey of recent Convolutional Neural Network (CNN) based approaches that have demonstrated significant improvements over earlier methods that rely largely on hand-crafted representations is provided.
- CNN-Based cascaded multi-task learning of high-level prior and density estimation for crowd counting543
A novel end-to-end cascaded network of CNNs to jointly learn crowd count classification and density map estimation achieves lower count error and better quality density maps as compared to the recent state-of-the-art methods.
- Deep Multitask Learning for Railway Track Inspection441
It is shown that detection performance can be improved by combining multiple detectors within a multitask learning framework, and this approach results in improved accuracy for detecting defects on railway ties and fasteners.
- 440
- C2AE: Class Conditioned Auto-Encoder for Open-Set Recognition428
An open-set recognition algorithm using class conditioned auto-encoders with novel training and testing methodologies is proposed and experiments show that the proposed method performs significantly better than the state of the art methods.
- 419
- Unsupervised Domain Adaptation of Object Detectors: A Survey351
This work describes in detail the domain adaptation problem for detection and presents an extensive survey of the various methods, and identifies multiple aspects of the problem that are most promising for future research.
- Uncertainty Guided Multi-Scale Residual Learning-Using a Cycle Spinning CNN for Single Image De-Raining349
The proposed Uncertainty guided Multi-scale Residual Learning (UMRL) network attempts to address the single image de-raining problem by learning the rain content at different scales and using them to estimate the final de-rained output.
- 335
- 330
- Unconstrained face verification using deep CNN features303
An algorithm for unconstrained face verification based on deep convolutional features and evaluate it on the newly released IARPA Janus Benchmark A (IJB-A) dataset as well as on the traditional Labeled Face in the Wild (LFW) dataset.
- 293
- KiU-Net: Overcomplete Convolutional Architectures for Biomedical Image and Volumetric Segmentation255
A new architecture for im- age segmentation- KiU-Net is designed which has two branches: an overcomplete convolutional network Kite-Net which learns to capture fine details and accurate edges of the input, and (2) U- net which learns high level features.
- 253
- MeGA-CDA: Memory Guided Attention for Category-Aware Unsupervised Domain Adaptive Object Detection242
This work attempts to incorporate category information into the domain adaptation process by proposing Memory Guided Attention for Category-Aware Domain Adaptation (MeGA-CDA), which consists of employing category-wise discriminators to ensure category-aware feature alignment for learning domain-invariant discriminative features.
- Ambiguous Medical Image Segmentation Using Diffusion Models230
A single diffusion model-based approach that produces multiple plausible outputs by learning a distribution over group insights that outperforms existing state-of-the-art ambiguous segmentation networks in terms of accuracy while preserving naturally occurring variation.
- 229
- 224
- Sparse Representation-Based Open Set Recognition217
A generalized Sparse Representation-based Classification (SRC) algorithm for open set recognition where not all classes presented during testing are known during training, which model the tail of those two error distributions using the statistical Extreme Value Theory (EVT).
- Image Fusion Transformer210
This work proposes a novel Image Fusion Transformer (IFT) where a transformer-based multi-scale fusion strategy that attends to both local and long-range information (or global context) in the image fusion process.
- Syn2Real Transfer Learning for Image Deraining Using Gaussian Processes209
This work proposes a Gaussian Process-based semi-supervised learning framework which enables the network in learning to derain using synthetic dataset while generalizing better using unlabeled real-world images.
- Multi-Level Bottom-Top and Top-Bottom Feature Fusion for Crowd Counting207
A network that involves a multi-level bottom-top and top-bottom fusion method to combine information from shallower to deeper layers and vice versa at multiple levels and a principled way of generating scale-aware ground-truth density maps for training.
- KiU-Net: Towards Accurate Segmentation of Biomedical Images Using Over-Complete Representations203
This work proposes an over-complete architecture (Ki-Net) which involves projecting the data onto higher dimensions (in the spatial sense) and results in significant improvements in the case of segmenting small anatomical landmarks and blurred noisy boundaries while obtaining better overall performance.
- 202
- Prior-Based Domain Adaptive Object Detection for Hazy and Rainy Conditions197
An unsupervised prior-based domain adversarial object detection framework for adapting the detectors to hazy and rainy conditions is proposed, which uses weather-specific prior knowledge obtained using the principles of image formation to define a novel prior-adversarial loss.
- 197
- HyperTransformer: A Textural and Spectral Feature Fusion Transformer for Pansharpening196
A novel attention mechanism for pansharpening called HyperTransformer is presented, in which features of LR-HSI and PAN are formulated as queries and keys in a transformer, respectively, which achieves significant improvement over the state-of-the-art methods on both spatial and spectral quality measures.
- Deep Multimodal Subspace Clustering Networks191
An affinity fusion-based network is also proposed in which the self-expressive layer corresponding to different modalities is enforced to be the same, in addition to various spatial fusion- based methods.
- Improving the Performance of Unimodal Dynamic Hand-Gesture Recognition With Multimodal Training184
This work presents an efficient approach for leveraging the knowledge from multiple modalities in training unimodal 3D convolutional neural networks (3D-CNNs) for the task of dynamic hand gesture recognition, and introduces a "spatiotemporal semantic alignment" loss (SSA) to align the content of the features from different networks.
- Multi-institutional Collaborations for Improving Deep Learning-based Magnetic Resonance Image Reconstruction Using Federated Learning182
A federated learning (FL) based solution in which the learned intermediate latent features among different source sites are aligned with the distribution of the latent features at the target site for MR image reconstruction.
- 179
- 177
- A deep pyramid Deformable Part Model for face detection170
The proposed method, DP2MFD, reduces the gap in training and testing of DPM on deep features by adding a normalization layer to the deep convolutional neural network (CNN).
- 165
- High-Quality Facial Photo-Sketch Synthesis Using Multi-Adversarial Networks153
A novel synthesis framework called Photo-Sketch Synthesis using Multi-Adversarial Networks, (PS2-MAN) that iteratively generates low resolution to high resolution images in an adversarial way that leverages the pair information using CycleGAN framework.
- Domain Adaptive Dictionary Learning151
A function learning framework for the task of transforming a dictionary learned from one visual domain to the other, while maintaining a domain-invariant sparse representation of a signal.
- Joint Transmission Map Estimation and Dehazing using Deep Networks144
This paper relax the constant atmospheric light assumption and proposes a novel unified single image dehazing network that jointly estimates the transmission map and performs dehaze and provides an end-to-end learning framework.
- ReconFormer: Accelerated MRI Reconstruction Using Recurrent Transformer143
A recurrent Transformer model, namely ReconFormer, for MRI reconstruction, which can iteratively reconstruct high-fidelity magnetic resonance images from highly under-sampled <inline-formula> <tex-math notation="LaTeX">${k}$ </tex-math></inline-formula>-space data, built upon Recurrent Pyramid Transformer Layers (RPTLs).
- 143
- 140
- Dictionary-based Face Recognition from Video | NIST140
This work introduces the concept of video-dictionaries for face recognition, which generalizes the work in sparse representation and dictionaries for faces in still images and performs significantly better than many competitive video-based face recognition algorithms.
- Dictionary-Based Face Recognition from Video140
This work introduces the concept of video-dictionaries for face recognition, which generalizes the work in sparse representation and dictionaries for faces in still images and performs significantly better than many competitive video-based face recognition algorithms.
- 136
- 134
- 130
- Learning from Ambiguously Labeled Face Images126
This work proposes a Matrix Completion for Ambiguity Resolution (MCar) method for predicting the actual labels from ambiguously labeled images, and extends MCar to incorporate the labeling constraints among instances when such prior knowledge is available.
- 122
- Instance Relation Graph Guided Source-Free Domain Adaptive Object Detection121
This paper proposes a novel training strategy for adapting a source-trained object detector to the target domain without source data, and designs a novel contrastive loss to enhance the target representations by exploiting the objects relations for a given target domain input.
- Revisiting Consistency Regularization for Semi-supervised Change Detection in Remote Sensing Images115
A semi-supervised CD model is proposed in which an unsupervisedCD loss is formulated in addition to the supervised Cross-Entropy (CE) loss by constraining the output change probability map of a given unlabeled bi-temporal image pair to be consistent under the small random perturbations applied on the deep feature difference map that is obtained by subtracting their latent feature representations.
- CLIP goes 3D: Leveraging Prompt Tuning for Language Grounded 3D Recognition114
This work proposes a new framework termed CG3D (CLIP Goes 3D) where a 3D encoder is learned to exhibit zero-shot capabilities and utilizes contrastive loss on 3D features obtained from the 3D encoder, as well as visual and text features extracted from CLIP.
- Deep Transfer Learning for Multiple Class Novelty Detection109
It is shown that thresholding the maximal activation of the proposed network can be used to identify novel objects effectively, and that the proposed method achieves significant improvements over the state-of-the-art methods.
- Pushing the Frontiers of Unconstrained Crowd Counting: New Dataset and Benchmark Method109
A novel crowd counting network that progressively generates crowd density maps via residual error estimation using VGG16 as the backbone network and employs density map generated by the final layer as a coarse prediction to refine and generate finer density maps in a progressive fashion using residual learning.
- 109
- An End-to-End System for Unconstrained Face Verification with Deep Convolutional Neural Networks103
An end-to-end system for the unconstrained face verification problem based on deep convolutional neural networks (DCNN) is presented and is evaluated using the newly released IARPA Janus Benchmark A (IJB-A) dataset and its extended version Janus Challenging set 2 (JANUS CS2) dataset.
- 102
- 101
- 100
- SAR Despeckling Using a Denoising Diffusion Probabilistic Model97
SAR-DDPM is introduced, a denoising diffusion probabilistic model for SAR despeckling that uses a Markov chain that transforms clean images into white Gaussian noise by successively adding random noise.
- Disentangled Variational Representation for Heterogeneous Face Recognition97
This paper model a face representation with an intrinsic identity information and its within-person variations and proposes an alternative optimization scheme for the disentangled variational representation part and the heterogeneous face recognition network part that effectively reduces the NIR and VIS domain discrepancy and alleviates over-fitting.
- 95
- 95
- Multimodal sparse and low-rank subspace clustering94
Experiments on face clustering indicate the proposed multimodal extensions of sparse subspace clustering and low-rank representation based sub space clustering algorithms for clustering data lying in a union of subspaces performs favorably compared to state-of-the-art sub Space clustering methods.
- Towards Zero-Shot Anomaly Detection and Reasoning with Multimodal Large Language Models93
Anomaly-OneVision (Anomaly-OV), the first specialist visual assistant for ZSAD and reasoning is proposed, which leverages a Look-Twice Feature Matching (LTFM) mechanism to adaptively select and emphasize abnormal visual tokens.
- 93
- 90
- 90
- Hyperspectral Pansharpening Based on Improved Deep Image Prior and Residual Reconstruction87
A novel over-complete network, called HyperKite, which focuses on learning high-level features by constraining the receptive from increasing in the deep layers is proposed, which demonstrates the superiority of the DIP-HyperKite over the state-of-the-art pansharpening methods.
- Synthesis of High-Quality Visible Faces from Polarimetric Thermal Faces using Generative Adversarial Networks84
A generative adversarial networks based multi-stream feature-level fusion technique to synthesize high-quality visible images from polarimetric thermal images and demonstrates that the proposed method achieves state-of-the-art performance.
- Pushing the Limits of Unconstrained Face Detection: a Challenge Dataset and Baseline Results84
This work identifies the next set of challenges that requires attention from the research community and collects a new dataset of face images that involve these issues such as weather-based degradations, motion blur, focus blur and several others.
- 83
- Learning to Count in the Crowd from Limited Labeled Data81
This work proposes a Gaussian Process-based iterative learning mechanism that involves estimation of pseudo-ground truth for the unlabeled data, which is then used as supervision for training the network.
- 80
- 80
- Generating high quality visible images from SAR images using CNNs79
Experimental results show that the proposed method can produce visible-like images better compared to the recent state-of-the-art deep learning-based methods.
- Transformer-Based SAR Image Despeckling78
The proposed despeckling network comprises of a transformer-based encoder which allows the network to learn global dependencies between different image regions - aiding in better despekling.
- Deblurring Face Images Using Uncertainty Guided Multi-Stream Semantic Networks78
The proposed Uncertainty Guided Multi-Stream Semantic Network (UMSN) processes regions belonging to each semantic class independently and learns to combine their outputs into the final deblurred result.
- Sparse Representations and Compressive Sensing for Imaging and Vision78
This book highlights some of the key mathematical insights underlying sparse representation and compressed sensing and illustrates the role of these theories in classical vision, imaging and biometrics problems.
- AdaMAE: Adaptive Masking for Efficient Spatiotemporal Learning with Masked Autoencoders77
AdaMAE, an adaptive masking strategy for MAEs that is end-to-end trainable that samples visible tokens based on the semantic context using an auxiliary sampling network, and shows that AdaMAE samples more tokens from the high spatiotemporal information regions, thereby allowing it to mask 95% of tokens.
- 76
- 75
- Generative adversarial network-based synthesis of visible faces from polarimetrie thermal faces73
This paper proposes a Generative Adversarial Network-based Visible Face Synthesis (GAN-VFS) method to synthesize more photo-realistic visible face images from their corresponding polarimetric images and demonstrates that the proposed method achieves state-of-the-art performance.
- 72
- Automatic segmentation of bone surfaces from ultrasound using a filter-layer-guided CNN71
The proposed convolutional neural network-based technique for segmentation of bone surfaces from in vivo US scans utilizes fusion of feature maps and employs multi-modal images to abate sensitivity to variations caused by imaging artifacts and low intensity bone boundaries.
- Towards Online Domain Adaptive Object Detection69
This work proposes a novel unified adaptation framework that adapts and improves generalization on the target domain in both offline and online settings for object detection and introduces MemXformer - a cross-attention transformer-based memory module where items in the memory take advantage of domain shifts and record prototypical patterns of the target distribution.
- Analyzing Bias in Diffusion-based Face Generation Models68
It is suggested that diffusion models tend to worsen distribution bias in the training data for various attributes, which is heavily influenced by the size of the dataset, while GAN models trained on balanced datasets with a larger number of samples show less bias across different attributes.
- Simultaneous Segmentation and Classification of Bone Surfaces from Ultrasound Using a Multi-feature Guided CNN68
A multi-feature guided convolutional neural network architecture is proposed for simultaneous enhancement, segmentation, and classification of bone surfaces from US data, validated on 650 in vivo US scans collected using two US machines.
- 68
- JeDi: Joint-Image Diffusion Models for Finetuning-Free Personalized Text-to-Image Generation66
Joint-Image Diffusion (JeDi) is proposed, an effective technique for learning a finetuning-free personalization model that achieves state-of-the-art gen-eration quality, both quantitatively and qualitatively, significantly outperforming both the prior finetuning-based and finetuning-free personalization baselines.
- View-decoupled Transformer for Person Re-identification under Aerial-ground Camera Network66
Experiments show that VDT is a feasible and effective solution for AGPReID, surpassing the previous method on mAP/Rank1 by up to 5.0%/2.7% on CARGO and 3.7%/5.2% on AG-ReID, keeping the same magnitude of computational complexity.
- AT-DDPM: Restoring Faces degraded by Atmospheric Turbulence using Denoising Diffusion Probabilistic Models65
This paper proposes the first DDPM-based solution for the problem of atmospheric turbulence mitigation and proposes a fast sampling technique for reducing the inference times for conditional DDPMs.
- Analysis sparse coding models for image-based classification65
It is demonstrated that the proposed co-sparse model for image classification is robust and efficient, while giving a comparable or better recognition performance than the traditional synthesis-based models.
- Facial attributes for active authentication on mobile devices64
This work trains a bunch of binary attribute classifiers which provide compact visual descriptions of faces which are able to capture meaningful attributes of faces and performs better than the previously proposed LBP-based authentication method.
- 63
- Sparse Embedding: A Framework for Sparsity Promoting Dimensionality Reduction63
An optimization problem for learning a transformation from the original signal domain to a lower-dimensional one in a way that preserves the sparse structure of data is formulated and an efficient optimization algorithm is proposed and its non-linear extension is presented.
- Distilling Multi-Modal Large Language Models for Autonomous Driving62
DiMA is an end-to-end autonomous driving system that maintains the efficiency of an LLM-free (or vision-based) planner while leveraging the world knowledge of an LLM, enabling robust planning without compromising on efficiency.
- SketchEdit: Mask-Free Local Image Manipulation with Partial Sketches62
This work investigates a new paradigm of sketch-based image manipulation: mask-free local image manipulation, which only requires sketch inputs from users and utilizes the entire original image.
- SceneComposer: Any-Level Semantic Image Synthesis61
Experimental results show that the proposed method can generate high-quality images following the layout at given precision, and compares favorably against existing methods.
- A Mamba-Based Siamese Network for Remote Sensing Change Detection60
A Mamba-based Change Detector (M-CD) that segments out the regions of interest even better and demonstrates lineartime training capabilities and an improved receptive field over transformers.
- 60
- Information-Theoretic Dictionary Learning for Image Classification60
This work presents a two-stage approach for learning dictionaries for object classification tasks based on the principle of information maximization, and seeks a dictionary that is compact, discriminative, and generative.
- Over-and-Under Complete Convolutional RNN for MRI Reconstruction59
This paper proposes an Over-and-Under Complete Convolutional Recurrent Neural Network (OUCR), which consists of an overcomplete and an undercomplete convolutional recurrent neural network that gives special attention in learning local structures by restraining the receptive field of the network.
- FaceXFormer: A Unified Transformer for Facial Analysis57
FaceXFormer is the first model capable of handling ten facial analysis tasks while maintaining real-time performance at 33.21 FPS, and introduces FaceX, a lightweight decoder with a novel bi-directional crossattention mechanism, which jointly processes face and task tokens to learn robust and generalized facial representations.
- DDPM-CD: Denoising Diffusion Probabilistic Models as Feature Extractors for Change Detection56
This change detection (CD) method proposes a novel way to incorporate the millions of off-the-shelf, unlabeled, remote sensing images available through different earth observation programs into the training process through denoising diffusion probabilistic models.
- Learning to Restore Images Degraded by Atmospheric Turbulence Using Uncertainty56
A deep learning-based approach for restring a single image degraded by atmospheric turbulence is proposed that makes use of the epistemic uncertainty based on Monte Carlo dropouts to capture regions in the image where the network is having hard time restoring.
- 56
- 55
- 55
- 54
- On-the-Fly Test-time Adaptation for Medical Image Segmentation54
A new framework called Adaptive UNet where each convolutional block is equipped with an adaptive batch normalization layer to adapt the features with respect to a domain code and validated the performance on both 2D and 3D data distribution shifts.
- A Large-Scale, Time-Synchronized Visible and Thermal Face Dataset54
This paper presents extensive benchmark results and analysis on thermal face landmark detection and thermal-to-visible face verification by evaluating state-of-the-art models on the ARL-VTF dataset.
- Unconstrained Still/Video-Based Face Verification with Deep Convolutional Neural Networks54
The design details of a deep learning system for unconstrained face recognition, including modules for face detection, association, alignment and face verification are presented and some open issues regarding DCNNs for face verification problems are discussed.
- 53
- 52
- Adversarially Robust One-class Novelty Detection51
The proposed method, referred to as Principal Latent Space (PrincipaLS), learns the incrementally-trained cascade principal components in the latent space to robustify novelty detectors and consistently enhances the adversarial robustness of novelty detection models.
- Anomaly Detection-Based Unknown Face Presentation Attack Detection51
A deep-learning solution for anomaly detection-based spoof attack detection where both classifier and feature representations are learned together end-to-end, which benefits from the representation learning power of the CNNs and learns better features for fPAD task as shown in the ablation study.
- 51
- 51
- Deep Learning for Cross-Domain Few-Shot Visual Recognition: A Survey50
This article presents the first comprehensive review of Cross-domain Few-shot Learning (CDFSL), a field that has received less attention compared to traditional FSL due to its unique challenges.
- CrowdDiff: Multi-Hypothesis Crowd Density Estimation Using Diffusion Models50
CrowdDiff, a conditional diffusion model that generates the crowd density map as a reverse diffusion process, out-performs existing state-of-the-art crowd counting methods on several public crowd analysis benchmarks with significant improvements.
- 50
- World-in-World: World Models in a Closed-Loop World49
This study curates four closed-loop environments that rigorously evaluate diverse WMs, prioritize task success as the primary metric, and move beyond the common focus on visual quality; it presents the first data scaling law for world models in embodied settings.
- Auto-FedRL: Federated Hyperparameter Optimization for Multi-institutional Medical Image Segmentation48
- SPIN Road Mapper: Extracting Roads from Aerial Images via Spatial and Interaction Space Graph Reasoning for Autonomous Driving48
A network based on stacked hourglass modules and SPIN pyramid for road segmentation which achieves better performance compared to existing methods and is computationally efficient and significantly boosts the convergence speed during training, making it feasible for applying on large-scale high-resolution aerial images.
- 48
- Latent Feature-Guided Diffusion Models for Shadow Removal47
This paper proposes the use of diffusion models as they offer a promising approach to gradually refine the details of shadow regions during the diffusion process, and improves this process by conditioning on a learned latent feature space that inherits the characteristics of shadow-free images, thus avoiding the limitation of conventional methods.
- Robust local features for remote face recognition47
Experimental results show promising results for two challenging datasets which have poor image quality, i.e., a remote face dataset and the Point and Shoot Face Recognition Challenge dataset.
- DeepMMSearch-R1: Empowering Multimodal LLMs in Multimodal Web Search46
DeepMMSearch-R1 is presented, the first multimodal LLM capable of performing on-demand, multi-turn web searches and dynamically crafting queries for both image and text search tools and introduces DeepMMSearchVQA, a novel multimodal VQA dataset created through an automated pipeline intermixed with real-world information from web search tools.
- STEREO: A Two-Stage Framework for Adversarially Robust Concept Erasing from Text-to-Image Diffusion Models46
STEREO is proposed, a novel two-stage framework that employs adversarial training as a first step rather than the only step for robust concept erasure and introduces an anchor-concept-based compositional objective to robustly erase the target concept in a single fine-tuning stage, while minimizing the degradation of model utility.
- Meta-UDA: Unsupervised Domain Adaptive Thermal Object Detection using Meta-Learning45
This work proposes an algorithm agnostic meta-learning framework to improve existing UDA methods instead of proposing a new UDA strategy, and demonstrates the superiority of the method over many baselines in the UDA setting, producing a state-of-the-art thermal detector for the KAIST and DSIAC datasets.
- Topical antiviral and antifungal medications in pregnancy: a review of safety profiles45
Recommendations on utilizing topical antiviral and antifungal medications in pregnancy are provided and Ketoconazole and selenium sulphide are likely safe, but should be employed in limited areas for brief periods.
- Remote identification of faces: Problems, prospects, and progress44
It is demonstrated that in addition to applying a good classification algorithm, finding features that are robust to variations mentioned above and developing statistical models which can account for these variations are very important for remote face recognition.
- Sparsity-motivated automatic target recognition44
An automatic target recognition algorithm using the recently developed theory of sparse representations and compressive sensing is presented and it is shown how sparsity can be helpful for efficient utilization of data for target recognition.
- Completely Self-supervised Crowd Counting via Distribution Matching42
This work proposes the proposed paradigm of complete self-supervision, which does not need even a single labeled image to be trained, and establishes the superiority of the method in less data setting as well.
- Open-Set Adversarial Defense42
An Open-Set Defense Network (OSDN) is proposed as a solution to the OSAD problem and an encoder with feature-denoising layers coupled with a classifier to learn a noise-free latent feature representation is employed.
- Discrete shearlet transform on GPU with applications in anomaly detection and denoising42
This paper presents an open source stand-alone implementation of the 2D discrete shearlet transform using CUDA C++ as well as GPU-accelerated MATLAB implementations of the2D and 3D shearlets transforms, and instrumented the code so that the running time of each kernel under different GPU hardware is analyzed.
- Deep Image Compositing41
A dense-connected multi-stream fusion network is proposed to effectively fuse the information from the foreground and back-ground images at different scales to automatically generate high-quality composites.
- Deep feature-based face detection on mobile devices41
The proposed deep feature-based face detector for mobile devices to detect user's face acquired by the front-facing camera is able to detect faces in images containing extreme pose and illumination variations as well as partial faces.
- 40
- 39
- 39
- CoDi: Conditional Diffusion Distillation for Higher-Fidelity and Faster Image Generation38
This work in-troduce a novel method, dubbed CoDi, that adapts a pre-trained latent diffusion model to accept additional image conditioning inputs while significantly reducing the sampling steps required to achieve high-quality results.
- Open-Set Adversarial Defense with Clean-Adversarial Mutual Learning38
This paper proposes an Open-Set Defense Network with Clean-Adversarial Mutual Learning (OSDN-CAML) as a solution to the OSAD problem and designs an encoder with dual-attentive feature-denoising layers coupled with a classifier to learn a noise-free latent feature representation, which adaptively removes adversarial noise.
- LTT-GAN: Looking Through Turbulence by Inverting GANs37
This work proposes a method in which well-trained Generative Adversarial Networks are leveraged as effective priors to mitigate the effect of turbulence, and proposes to connect hierarchical layers of a GAN to get multiple images with the same identity at the same time, which enriches the context and helps to better improve the restoration performance.
- Federated Generalized Face Presentation Attack Detection37
A federated face presentation attack detection framework that simultaneously takes advantage of rich fPAD information available at different data owners while preserving data privacy is proposed and a federated domain disentangling strategy is introduced in FedGPAD.
- Active Authentication using an Autoencoder regularized CNN-based One-Class Classifier37
A convolutional neural network (CNN) based approach for one class classification in which a zero centered Gaussian noise and an autoencoder are used to model the pseudo-negative class and to regularize the network to learn meaningful feature representations for oneclass data, respectively.
- 37
- Domain Adaptation for Visual Recognition37
This monograph provides a comprehensive view of domain adaptation solutions for visual recognition problems under uncertainty, analyzing the challenges posed by the realm of "big visual data", in terms of the generalization ability of adaptation algorithms to unconstrained data acquisition as well as issues related to their computational tractability.
- 37
- 37
- Think Before You Diffuse: Infusing Physical Rules into Video Diffusion36
This work proposes DiffPhy, a generic framework that enables physically-correct and photo-realistic video generation by fine-tuning a pre-trained video diffusion model and leverages large language models (LLMs) to explicitly reason a comprehensive physical context from the text prompt and use it to guide the generation.
- 36
- 36
- T2V-DDPM: Thermal to Visible Face Translation using Denoising Diffusion Probabilistic Models35
This paper proposes a novel inference strategy for speeding up the inference time of DDPMs, specifically for the problem of T2V image translation and achieves the state-of-the-art results on multiple datasets.
- Exploring Overcomplete Representations for Single Image Deraining Using CNNs35
This article proposes using an overcomplete convolutional network architecture which gives special attention in learning local structures by restraining the receptive field of filters and combines it with U-Net so that it does not lose out on the global structures while focusing more on low-level features, to compute the derained image.
- Learning to Segment Brain Anatomy From 2D Ultrasound With Less Data35
A novel image synthesis method using multi-scale self attention generator to synthesize high-quality US images from various segmentation masks and introduces a technique which guides CBAS to learn the weights based on the confidence measure about the estimate.
- Inverse Attention Guided Deep Crowd Counting Network35
This paper presents Inverse Attention Guided Deep Crowd Counting Network (IA-DCCN), a single-step training framework that efficiently infuses segmentation information through an inverse attention mechanism into the counting network, resulting in significant improvements.
- Polarimetric Thermal to Visible Face Verification via Attribute Preserved Synthesis35
A novel Attribute Preserved Generative Adversarial Network is proposed to synthesize the visible image from the thermal image guided by the extracted attributes, which achieves significant improvements over the state-of-the-art methods.
- 35
- Convolutional Sparse Coding-based Image Decomposition35
This work proposes a sparsity-based optimization framework to decompose a given image into cartoon and texture components by working directly on the whole image, and shows that the proposed method performs favorably compared to state-of-the-art image separation methods.
- 34
- FaceXBench: Evaluating Multimodal LLMs on Face Understanding33
This work introduces FaceXBench, a comprehensive benchmark designed to evaluate MLLMs on complex face understanding tasks, revealing the unique challenges in complex face understanding tasks.
- Defending Against Multiple and Unforeseen Adversarial Videos33
This paper proposes one of the first defense strategies against multiple types of adversarial videos for video recognition, and exhibits stronger multi-perturbation robustness against different and even unforeseen adversarial video types, ranging from Lp-bounded attacks and physically realizable attacks.
- 33
- 33
- 32
- In2I: Unsupervised Multi-Image-to-Image Translation Using Generative Adversarial Networks32
This paper introduces a Generative Adversarial Network (GAN) based framework along with a multi-modal generator structure and a new loss term, latent consistency loss, and shows that leveraging multiple inputs generally improves the visual quality of the translated images.
- 32
- 32
- Joint Sparsity-Based Robust Multimodal Biometrics Recognition32
A novel multimodal multivariate sparse representation method for multi-modal biometrics recognition, which represents the test data by a sparse linear combination of training data, while constraining the observations from different modalities of the test subject to share their sparse representations.
- 31
- Bigger is not Always Better: Scaling Properties of Latent Diffusion Models29
An in-depth investigation into how model size influences sampling efficiency across varying sampling steps unveil a surprising trend: when operating under a given inference budget, smaller models frequently outperform their larger equivalents in generating high-quality results.
- 29
- Semi-Supervised Image Deraining Using Gaussian Processes29
This work proposes a Gaussian Process-based semi-supervised learning framework which enables the network in learning to derain using synthetic dataset while generalizing better using unlabeled real-world images.
- Generalized Dictionaries for Multiple Instance Learning29
This work presents a multi-class multiple instance learning (MIL) algorithm using the dictionary learning framework where the data is given in the form of bags and proposes a noisy-OR model and a generalized mean-based optimization framework for learning the dictionaries in the feature space.
- 29
- Disruptive Autoencoders: Leveraging Low-level features for 3D Medical Image Pre-training28
Disruptive Autoencoders is introduced, a pre-training framework that attempts to reconstruct the original image from disruptions created by a combination of local masking and low-level perturbations.
- 28
- Target and Task specific Source-Free Domain Adaptive Image Segmentation28
A systematic two-stage approach for SFUDA comprising of target- specific adaptation followed by task-specific adaptation, which enhances the pseudo-label generation by minimizing high entropy regions using the proposed ensemble entropy minimization loss and a selective voting strategy.
- 28
- Mask-Free OVIS: Open-Vocabulary Instance Segmentation without Manual Mask Annotations27
This work learns both base and novel categories from pseudo-mask annotations generated by the vision-language model in a weakly supervised manner using the proposed Mask-free OVIS pipeline, which significantly improves the mAP scores on the MS-COCO dataset and OpenImages dataset.
- Federated Learning-based Active Authentication on Mobile Devices27
This work proposes a novel method that is able to tackle heterogeneous/non-IID distribution of data in FAA and shows that such approach performs better than state-of-the-art one-class based FAA methods and is also able to outperform traditional FL/SL methods.
- Anatomic and Molecular MR Image Synthesis Using Confidence Guided CNNs27
A confidence-guided SAMR (CG-SAMR) that synthesizes data from lesion contour information to multi-modal MR images, including T1-weighted (T1w), gadolinium enhanced T1w (Gd-T1W), T2-weighting (T2w), and fluid-attenuated inversion recovery (FLAIR), as well as the molecular amide proton transferweighted sequence.
- 27
- Dictionary-Based Face and Person Recognition From Unconstrained Video27
It is shown that fusing face and body identity cues can improve performance over face alone, and the experimental results compare favorably with those published in the literature.
- 26
- 26
- 26
- Bone shadow segmentation from ultrasound data for orthopedic surgery using GAN26
This work proposes a computational method, based on a novel generative adversarial network (GAN) architecture, to segment bone shadow images from in vivo US scans in real-time and shows how these segmented shadow images can be incorporated, as a proxy, to a multi-feature guided convolutional neural network (CNN) architecture for real- time and accurate bone surface segmentation.
- 26
- Non-linear dictionary learning with partially labeled data26
This work proposes a non-linear discriminative dictionary learning technique which utilizes both labeled and unlabeled data for learning dictionaries in the high-dimensional feature space and shows how this method can be extended for ambiguously labeled classification problem where each training sample has multiple labels and only one of them is correct.
- Robust real-time bone surfaces segmentation from ultrasound using a local phase tensor-guided CNN25
An intensity-invariant convolutional neural network architecture is proposed for robust segmentation of bone surfaces from US data obtained from two different US machines with varying acquisition settings and encouraging results suggest that the proposed method is promising enough for further evaluation.
- MambaRecon: MRI Reconstruction with Structured State Space Models24
This work proposes an inno-vative MRI reconstruction framework that employs structured state space models at its core, aimed at amplifying both long-range contextual sensitivity and reconstruction efficacy.
- 24
- Steered Diffusion: A Generalized Framework for Plug-and-Play Conditional Image Synthesis24
Steered Diffusion is introduced, a generalized framework for photorealistic zero-shot conditional image generation using a diffusion model trained for unconditional generation by designing a loss using a pre-trained inverse model that characterizes the conditional task.
- Design of Non-Linear Discriminative Dictionaries for Image Classification24
This work proposes a kernel driven simultaneous orthogonal matching pursuit algorithm for the task of sparse coding in the feature space and demonstrates that the proposed non-linear dictionary learning method is robust and can perform significantly better than many competitive discriminative dictionary learning algorithms.
- Overcomplete Deep Subspace Clustering Networks23
This work proposes a simple yet efficient alternative method-Overcomplete Deep Subspace Clustering Networks (ODSC) where the features from both undercomplete and overcomplete autoencoder networks are fuse before passing them through the self-expressive layer thus enabling us to extract a more meaningful and robust representation of the input data for clustering.
- 23
- 23
- 23
- Deep Metric Learning for Unsupervised Remote Sensing Change Detection22
An unsupervised CD method based on deep metric learning that achieves significant improvements over the state-of-the-art supervised and unsupervised CD methods.
- Directional Multiscale Processing of Images Using Wavelets with Composite Dilations22
This work introduces and applies a novel multiscale image decomposition algorithm for the efficient digital implementation of wavelets with composite dilations, and provides consistent improvements upon competing state-of-the-art methods.
- MixedNUTS: Training-Free Accuracy-Robustness Balance via Nonlinearly Mixed Classifiers21
MixedNUTS is proposed, a training-free method where the output logits of a robust classifier and a standard non-robust classifier are processed by nonlinear transformations with only three parameters, which are optimized through an efficient algorithm.
- Spatio-Temporal Pixel-Level Contrastive Learning-based Source-Free Domain Adaptation for Video Semantic Segmentation21
Spatio-Temporal Pixel-Level (STPL) contrastive learning is proposed, a novel method that takes full advantage of spatio-temporal information to tackle the absence of source data better than current UDA and SFDA approaches.
- Deep-learning-enabled brain hemodynamic mapping using resting-state fMRI21
It is shown that rs-fMRI can be used to map cerebral hemodynamic function and delineate impairment, and deep learning enables reproducible mapping of cerebrovascular reactivity and bolus arrival time of the human brain using resting-state CO_2 fluctuations as a natural “contrast media”.
- A Synthesis-Based Approach for Thermal-to-Visible Face Verification21
This paper presents an algorithm that achieves state-of-the-art performance on both the ARL-VTF and TUFTS multi-spectral face datasets, and shows that the proposed method significantly outperforms face frontalization methods on profile-to-frontal verification.
- Dictionaries for image and video-based face recognition [Invited]21
The role of sparse representation and dictionary learning for efficient face identification and verification is reviewed and discriminative dictionary learning algorithms as well as methods based on weakly supervised learning and domain adaptation are summarized.
- Reference-based Controllable Scene Stylization with Gaussian Splatting20
ReGS is proposed, which adapts 3D Gaussian Splatting (3DGS) for reference-based stylization to enable real-time stylized view synthesis and can produce state-of-the-art stylization results that respect the reference texture while embracing real-time rendering speed for free-view navigation.
- Securing Deep Generative Models with Universal Adversarial Signature20
This paper proposes to inject a universal adversarial signature into an arbitrary pre-trained generative model, in order to make its generated contents more detectable and traceable.
- Enhancing Adversarial Robustness for Deep Metric Learning20
Comprehensive experimental results suggest that the proposed Hardness Manipulation, although simple in its form, overwhelmingly outperforms the state-of-the-art de-fenses in terms of robustness, training efficiency, as well as performance on benign examples.
- 20
- The Power of Context: How Multimodality Improves Image Super-Resolution19
This work proposes a novel approach that leverages the rich contextual information available in multiple modalities to learn a powerful generative prior for SISR within a diffusion model framework, and mitigate hallucinations, often introduced by text prompts, by using spatial information from other modalities to guide regional text-based conditioning.
- Fine-Context Shadow Detection using Shadow Removal19
This work proposes a Fine Context-aware Shadow Detection Network (FCSD- Net), and proposes a new learning strategy, called Restore to Detect (R2D), where it is shown that when a deep neural network is trained for restoration (shadow removal), it learns meaningful features to delineate the shadow masks as well.
- Evaluation of the Robustness of Learned MR Image Reconstruction to Systematic Deviations Between Training and Test Data for the Models from the fastMRI Challenge19
This work tested the generalization performance of the submissions with respect to various perturbations, and despite differences in model architecture and training, all of the methods perform very similarly.
- GAN-Based Realistic Bone Ultrasound Image and Label Synthesis for Improved Segmentation19
This work proposes a computational method, based on a novel generative adversarial network (GAN) architecture, to produce synthetic B-mode US images and their corresponding segmented bone surface masks in real-time and shows how a duality concept can be implemented for such tasks.
- 19
- 19
- Multiple kernel-based dictionary learning for weakly supervised classification19
A multiple instance learning (MIL) algorithm using the dictionary learning framework where the labels are given in the form of positive and negative bags, with each bag containing multiple samples to handle weakly supervised data.
- 19
- 19
- GenDeg: Diffusion-based Degradation Synthesis for Generalizable All-In-One Image Restoration18
GenDeg, a degradation and intensity-aware conditional diffusion model, capable of producing diverse degradation patterns on clean images, is introduced, capable of producing diverse degradation patterns on clean images.
- Active Learning for Vision-Language Models18
This work proposes a novel active learning (AL) framework that enhances the zero-shot classification performance of VLMs by selecting only a few informative samples from the unlabeled data for annotation during training.
- 18
- DAFE-FD: Density Aware Feature Enrichment for Face Detection18
This work approaches the problem of small face detection with the motivation of enriching the feature maps using a density map estimation module, inspired by recent crowd counting/density estimation techniques.
- 18
- 18
- Towards the design of an end-to-end automated system for image and video-based recognition18
The design details of a deep learning system for end-to-end unconstrained face verification/recognition and some open issues regarding DCNNs for object recognition problems are discussed.
- 18
- 18
- 18
- SAR Despeckling Using Overcomplete Convolutional Networks17
The proposed network consists of an overcomplete branch to focus on the local structures and an undercomplete branch that focuses on the global structures that improves despeckling performance compared to recent despekling methods on synthetic and real SAR images.
- Learning Feature Decomposition for Domain Adaptive Monocular Depth Estimation17
A novel UDA method for MDE, referred to as Learning Feature Decomposition for Adaptation (LFDA), which learns to decompose the feature space into content and style components, which achieves superior accuracy and lower computational cost compared to the state-of-the-art approaches.
- Multiple Class Novelty Detection Under Data Distribution Shift17
This work considers the problem of multiple class novelty detection under dataset distribution shift and proposes a method which brings together the techniques from novelty detection and domain adaptation to improve generalization of multiple class novelty detection on different domains.
- Sequential Score Adaptation with Extreme Value Theory for Robust Railway Track Inspection17
This paper uses extreme value theory (EVT) within a Bayesian framework to optimally adjust the sensitivity of anomaly detectors by approximating the lower tail of the probability density function of the scores with an Exponential distribution and using the Gamma conjugate prior learned from the training data.
- Attentive Prototypes for Source-free Unsupervised Domain Adaptive 3D Object Detection16
This work proposes a single-frame approach for source-free, un-supervised domain adaptation of LiDAR-based 3D object detectors that uses class prototypes to mitigate the effect of pseudo-label noise and shows that this method outperforms recent source-free domain adaptation works as well as those that leverage source information during training.
- 16
- 16
- Federated Test-Time Adaptive Face Presentation Attack Detection with Dual-Phase Privacy Preservation16
A Federated Test-Time Adaptive Face Presentation Attack Detection with Dual-Phase Privacy Preservation framework is proposed, with the aim of enhancing the generalization ability of fPAD models in both training and testing phase while preserving data privacy.
- 16
- 16
- Filter Images First, Generate Instructions Later: Pre-Instruction Data Selection for Visual Instruction Tuning15
Pre-Instruction Data Selection (PreSel), a more practical data selection paradigm that directly selects the most beneficial unlabeled images and generates instructions only for the selected images, achieves performance comparable to full-data VIT on the LLaVA-1.5 and Vision-Flan datasets.
- 3SD: Self-Supervised Saliency Detection With No Labels15
Experiments on six benchmark datasets demonstrate that without any labels, the 3SD method outperforms all existing weakly supervised and unsupervised methods, and its performance is on par with the fully-supervised methods.
- 15
- Heterogeneous Face Frontalization via Domain Agnostic Learning15
A domain agnostic learning-based generative adversarial network (DAL-GAN) which can synthesize frontal views in the visible domain from thermal faces with pose variations and can generate better quality frontal views compared to the other baseline methods.
- 15
- 15
- Wonder: Video World Model Done Better14
This work introduces a novel camera conditioning with a dense coordinate field whose renderings provide spatially aligned motion and orientation cues, allowing the model to interpret camera motion directly as visual evidence, regardless of actual context length.
- Open-Set Automatic Target Recognition14
This work proposes an Open-set Automatic Target Recognition framework where it enables open-set recognition capability for ATR algorithms and introduces a plugin Category-aware Binary Classifier module to effectively tackle unknown classes seen during inference.
- Lesion Mask-Based Simultaneous Synthesis of Anatomic and Molecular MR Images Using a GAN14
A method, called synthesis of anatomic and molecular MR images network (SAMR), which can simultaneously synthesize data from arbitrary manipulated lesion information on multiple anatomic and molecular MRI sequences, including T1-weighted (T1w), gadolinium enhanced T1w (Gd-T1w), T2-weighted (T2w), fluid-attenuated inversion recovery (FLAIR), and amide proton transfer-weighted (APTw).
- 14
- Few Is Enough: Task-Augmented Active Meta-learning for Brain Cell Classification14
It is shown that the proposed task-augmented meta-learning framework can learn to classify new cell types after a single gradient step with a limited number of training samples, and that active learning with Bayesian uncertainty can further improve the performance when the number ofTraining samples is extremely small.
- Improving Amide Proton Transfer-Weighted MRI Reconstruction Using T2-Weighted Images14
This paper presents a novel APTw image reconstruction framework that can accelerate APTw imaging by reconstructing APTw images directly from highly undersampled k-space data and corresponding T2w image at the same location and evaluates the method on two real datasets consisting of brain data from rats and humans.
- Domain Adaptive Subspace Clustering14
A general method that learns the projections of data in a space where the sparsity or low-rankness of data is maintained and an efficient iterative procedure for solving the proposed optimization problems is proposed.
- Toward a non-intrusive, physio- behavioral biometric for smartphones14
The novelty of the approach is to combine 3D gesture and face recognition in a nonintrusive and unconstrained environment; the active authentication system is running in the background while the user is performing his/her main task.
- Dictionary-based multiple instance learning14
This work presents a multi-class, multiple instance learning (MIL) algorithm using the dictionary learning framework where the data is given in the form of bags, and proposes a noisy-OR model-based optimization framework for learning the dictionaries.
- Guarding Barlow Twins Against Overfitting with Mixed Samples13
Mixed Barlow Twins is introduced, which aims to improve sample interaction during Barlow Twins training via linearly interpolated samples, and effectively mitigates feature over-fitting and further enhances the downstream performance on CIFAR-10, CIFAR-100, TinyImageNet, STL-10, and ImageNet datasets.
- Exploring Adversarially Robust Training for Unsupervised Domain Adaptation13
This paper provides a systematic study into multiple AT variants that can be applied to UDA and proposes a novel Adversarially Robust Training method, referred to as ARTUDA, which consistently improves the adversarial robustness of UDA models.
- 13
- 13
- 13
- 13
- Frame by Familiar Frame: Understanding Replication in Video Diffusion Models12
A systematic investigation into the phenomenon of sample replication in video diffusion models is presented, and new evaluation strategies that take replication into account are proposed, offering a more accurate measure of a model's ability to generate the original content.
- 12
- Hyp-OC: Hyperbolic One Class Classification for Face Anti-Spoofing12
This paper reformulates the face anti-spoofing task from a one-class perspective and proposes a novel hyperbolic one-class classification framework, believed to be the first work extending hyperbolic embeddings for face anti-spoofing in a one-class manner.
- 12
- PosSAM: Panoptic Open-vocabulary Segment Anything11
An open-vocabulary panoptic segmentation model that effectively unifies the strengths of the Segment Anything Model with the vision-language CLIP model in an end-to-end framework and proposes a novel Local Discriminative Pooling (LDP) module leveraging class-agnostic SAM and class-aware CLIP features for unbiased open-vocabulary classification.
- GLSFormer: Gated - Long, Short Sequence Transformer for Step Recognition in Surgical Videos11
This paper proposes a vision transformer-based approach to jointly learn spatio-temporal features directly from sequence of frame-level patches that intelligently combines short-term and long-term spatio -temporal feature representations.
- 11
- A Comparison of Different Atmospheric Turbulence Simulation Methods for Image Restoration11
This paper systematically evaluates the effectiveness of various turbulence simulation methods on image restoration and provides guidance to the researchers and practitioners working in this field to choose the suitable data generation models for training deep models for turbulence mitigation.
- Overcomplete Representations Against Adversarial Videos11
Experimental results show that the defenses focusing on images may be ineffective to videos, while OUDefend enhances robustness against different types of adversarial videos, ranging from additive attacks, multiplicative attacks to physically realizable attacks.
- Multimodal Face Synthesis From Visual Attributes11
A novel generative adversarial network which simultaneously synthesizes identity preserving multimodal face images (i.e., visible, sketch, thermal, etc.) from visual attributes without requiring paired data in different domains for training the network.
- 11
- 11
- 11
- 10
- Multimodal 3D Object Detection on Unseen Domains10
A multimodal fusion and supervised contrastive learning framework for 3D object detection that performs alignment of object features from sameclass samples of different domains while pushing the features from different classes apart yields state-of-the-art domain generalization performance under multiple dataset shifts.
- PETALface: Parameter Efficient Transfer Learning for Low-Resolution Face Recognition10
PETALface is the first work leveraging the powers of PEFT for low resolution face recognition, and extensive experiments demonstrate that the proposed method outperforms full fine-tuning on low-resolution datasets while preserving performance on high-resolution and mixed-quality datasets, all while using only 0.48% of the parameters.
- Shape-guided Object Inpainting10
A new data preparation method and a novel Contextual Object Generator (CogNet) for the object inpainting task are proposed that can generate realistic objects that fit the context in terms of both visual appearance and semantic meanings.
- MultAV: Multiplicative Adversarial Videos10
A novel attack method against video recognition models, Multiplicative Adversarial Videos (MultAV), which imposes perturbation on video data by multiplication, and can be generalized to not only $\ell_{p}$-norm attacks with a new adversary constraint called ratio bound, but also different types of physically realizable attacks.
- Quickest Intruder Detection For Multiple User Active Authentication10
The Multiple-user Quickest Intruder Detection (MQID) algorithm is formulated and extended to the data-efficient scenario where intruder detection is carried out with fewer observation samples.
- Deep Learning for Tattoo Recognition10
Deep convolutional neural network-based methods for automatic matching of tattoo images based on the AlexNet and Siamese networks are presented and it is shown that rather than using a simple contrastive loss function, triplet loss function can significantly improve the performance of a tattoo matching system.
- 10
- Sparsity inspired automatic target recognition10
This paper develops a framework for using only the needed data for automatic target recognition (ATR) algorithms using the recently developed theory of sparse representations and compressive sensing and shows how sparsity can be helpful for efficient utilization of data.
- Inverse halftoning using a shearlet representation10
The performance of the method improves upon many of the state-of-the-art inverse halftoning routines, including a wavelet-based method and a method that shares some similarities to a shearlet-type decomposition known as the local polynomial approximation (LPA) technique.
- Referring Change Detection in Remote Sensing Imagery9
A two-stage framework consisting of RCDNet, a cross-modal fusion network designed for referring change detection, and RCDGen, a diffusion-based synthetic data generation pipeline that produces realistic post-change images and change maps for a specified category using only pre-change image, without relying on semantic segmentation masks and thereby significantly lowering the barrier to scalable data creation is proposed.
- 9
- 9
- 9
- 9
- Screen-based active user authentication9
Preliminary results indicate that further research in using screen-based recordings for active authentication can lead to a reliable soft cyber biometric.
- On-Policy Distillation with Best-of-N Teacher Rollout Selection8
BRTS, a Best-of-N Rollout Teacher Selection framework for on-policy distillation, augments standard student-context OPD with a teacher-context supervision branch constructed from the curated teacher trajectory, which improves over standard OPD on challenging reasoning benchmarks.
- 8
- Interactive Portrait Harmonization8
This work introduces interactive harmonization, a new setting where the harmonization is performed with respect to a selected region in the reference image instead of the entire background, and introduces a new luminance matching loss to optimally match the color/luminance conditions between the composite foreground and select reference region.
- Bi-Noising Diffusion: Towards Conditional Diffusion Models with Generative Restoration Priors8
A new method is introduced that brings the predicted samples to the training data manifold using a pretrained unconditional diffusion model that acts as a regularizer and reduces the divergence introduced by the conditional model at each sampling step.
- Image Generation with Multimodal Priors using Denoising Diffusion Probabilistic Models8
This work proposes a solution based on a denoising diffusion probabilistic models to synthesise images under multi-model priors that does not require explicit retraining for all modalities and can leverage the outputs of individual modalities to generate realistic images according to different constraints.
- Utilizing Patch-Level Category Activation Patterns for Multiple Class Novelty Detection8
This paper proposes a novel method that makes deep convolutional neural networks robust to novel classes, and shows that the proposed method is able to identify novel class samples better compared to the other deep convolutional neural network-based methods.
- 8
- FreeViS: Training-free Video Stylization with Inconsistent References7
This paper proposes FreeViS, a training-free video stylization framework that generates stylized videos with rich style details and strong temporal coherence, and integrates multiple stylized references to a pretrained image-to-video (I2V) model, effectively mitigating the propagation errors observed in prior works.
- 7
- 7
- 7
- 7
- 7
- 7
- 7
- 7
- An M-channel directional filter bank compatible with the contourlet and shearlet frequency tiling7
New methods for creating M-channel directional filters to construct multiresolution and multidirectional orthogonal/biorthogonal transforms are presented, capable of creating directional filters that yield spatial-frequency tilings equivalent to those of the contourlet and the shearlet transforms.
- 6
- 6
- 6
- Attention Prompt Tuning: Parameter-efficient Adaptation of Pre-trained Models for Spatiotemporal Modeling6
Attention Prompt Tuning is introduced - a computationally efficient variant of prompt tuning for video-based applications such as action recognition that greatly reduces the number of FLOPs and latency while achieving a significant performance boost over the existing parameter-efficient tuning methods on UCF101, HMDB51, and SSv2 datasets for action recognition.
- 6
- 6
- 6
- Thermal to Visible Image Synthesis Under Atmospheric Turbulence6
This is the first work that studies the problem of thermal to visible image translation under atmospheric turbulence and an end-to-end reconstruction method is proposed which can directly transform thermal images into visible-spectrum images by utilizing natural image priors based on a pre-trained StyleGAN2 network.
- 6
- Federated Face Anti-spoofing.6
The experimental setting is introduced to evaluate the proposed FedFAS framework and extensive experiments are carried out to provide various insights about federated learning for FAS.
- 6
- Improving the Security of SSO in Distributed Computer Network using Digital Certificate and one Time Password (OTP)6
This work analyzes security flaws in single sign on, a new authentication mechanism for user to use multiple services provided by service provider in distributed computer network, and proposes solution for those flaws.
- Robust and Secure Iris Recognition6
This chapter discusses a unified framework based on sparse representations and random projections that can address the challenges of unconstrained acquisition, privacy enhancement without compromising security, and robust matching.
- 5
- ReBotNet: Fast Real-Time Video Enhancement5
This work designs an efficient and fast framework to perform real-time video enhancement for practical use-cases like live video calls and video streams and shows extensive results on multiple datasets where ReBot-Net outperforms existing approaches with lower computations, reduced memory requirements, and faster inference time.
- 5
- SyncNoise: Geometrically Consistent Noise Prediction for Instruction-based 3D Editing5
SyncNoise synchronously edits multiple views with 2D diffusion models while enforcing multi-view noise predictions to be geometrically consistent, which ensures global consistency in both semantic structure and low-frequency appearance.
- 5
- 5
- Towards Federated Learning Under Resource Constraints via Layer-wise Training and Depth Dropout5
This work introduces a simple yet effective strategy, Federated Layer-wise Learning, to simultaneously reduce per-client memory, computation, and communication costs and demonstrates that performance in downstream tasks is comparable to conventional federated self-supervised learning.
- Unsupervised Restoration of Weather-affected Images using Deep Gaussian Process-based CycleGAN5
This work introduces new losses for training CycleGAN that lead to more effective training, resulting in high quality reconstructions, and enables the CycleGAN architecture to transfer the knowledge from one domain (weather-degraded) to another (clean) more effectively.
- Error Diffusion Halftoning Against Adversarial Examples5
Experimental results demonstrate that the proposed method is able to improve adversarial robustness even under advanced adaptive attacks, while most of the other image transformation-based defenses do not.
- 5
- 5
- Robust Bone Shadow Segmentation from 2D Ultrasound Through Task Decomposition5
This work decomposes the bone shadow segmentation task into two subtasks, coarse bone shadow enhancement (BSE) and horizontal bone interval mask (HBIM) estimation and proposes a conditional shape discriminator to regularize the shape of the output segmentation map.
- 5
- Spine Surface Segmentation from Ultrasound Using Multi-feature Guided CNN5
A convolutional neural network (CNN)-based technique for segmenting spine surfaces from in vivo US scans that utilizes fusion of feature maps extracted from multimodal images to abate sensitivity to variations caused by imaging artifacts and low intensity bone boundaries.
- 5
- Fps-Sft: A Multi-Dimensional Sparse Fourier Transform Based on the Fourier Projection-Slice Theorem5
A sparse image reconstruction application is illustrated, which shows the capability of the FPS-SFT in solving less sparse scenarios containing non-uniformly distributed frequencies.
- 5
- Computational LADAR imaging5
A prototype of such a sensor having the capability to generate arbitrary illumination beam patterns rather than spot, line scanning, or flash techniques is described along with computational imaging algorithms to reduce speckle and identify scene objects in a low-dimensional compressed basis rather than in the pixel basis.
- Salient views and view-dependent dictionaries for object recognition5
A sparse representation-based approach to determine the salient views of 3D objects and the notion of view-dependent dictionaries built from salient views for applications in 3D object recognition and retrieval is introduced.
- 5
- 5
- DiffRegCD: Integrated Registration and Change Detection with Diffusion Features4
Experiments show that DiffRegCD outperforms recent baselines and remains robust under wide temporal and viewpoint variation, establishing diffusion features and classification-based correspondence as a strong foundation for integrated CD.
- Endless World: Real-Time 3D-Aware Long Video Generation4
To support infinite video generation, a conditional autoregressive training strategy is introduced that aligns newly generated content with existing video frames, preserving long-range dependencies while remaining computationally efficient, enabling real-time inference on a single GPU without additional training overhead.
- Evaluation of UNeXt for Automatic Bone Surface Segmentation on Ultrasound Imaging in Image-Guided Pediatric Surgery4
The segmentation model performs sufficiently across different pediatric age groups, making it a viable tool for integration into ultrasound-based patient registration in image-guided pediatric surgery.
- Harmonyseg: Tubular Structure Segmentation With Deep-Shallow Feature Fusion and Growth-Suppression Balanced Loss4
A new tubular structure segmentation framework named HarmonySeg is proposed, featuring a deep-to-shallow decoder network featuring flexible convolution blocks with varying receptive fields that can accurately segment 2D and 3D tubular structures and outperform existing state-of-the-art methods.
- 4
- Reference-Guided Identity Preserving Face Restoration4
A novel approach that maximizes reference face utility for improved face restoration and identity preservation and achieves state-of-the-art identity preserving restoration on benchmarks such as FFHQ-Ref and CelebA-Ref-Test, consistently outperforming previous work.
- Scaling Transformer-Based Novel View Synthesis with Models Token Disentanglement and Synthetic Data4
This work proposes a token disentanglement process within the transformer architecture, enhancing feature separation and ensuring more effective learning, which improves re-construction quality over standard transformers and enables scalable training with synthetic data.
- Attention Prompt Tuning: Parameter-efficient Adaptation of Pre-trained Models for Action Recognition4
- Dreamguider: Improved Training free Diffusion-based Conditional Generation4
This work proposes Dreamguider, a method that enables inference-time guidance without compute-heavy backpropagation through the diffusion network, and proposes an empirical guidance scale that works for a wide variety of tasks, hence removing the need for handcrafted parameter tuning.
- Behavior analysis of search and rescue operations employing human-machine teaming4
During the summer of 2022, the United States Military Academy hosted a robotics apprenticeship program during which interns programmed maritime robots to move autonomously to compete against each other in a force-on-force competition based on capture-the-flag game rules.
- JR2Net: Joint Monocular 3D Face Reconstruction and Reenactment4
This work proposes a novel cascade framework named JR2Net for Joint Face Reconstruction and Reenactment, which begins with the training of a coarse reconstruction network, followed by a 3D-aware face reenactments network based on the coarse reconstruction results, and trains an expression tracking network which can further enhance the coarse face reconstruction.
- Image Inpainting with Contextual Reconstruction Loss4
Experimental results demonstrate that the proposed inpainting model with the CR loss compares favourably against the state-of-the-arts in terms of quantitative and visual performance.
- Exploring Body Texture From mmW Images for Person Recognition4
Different multi-algorithmic and multi-modal techniques are analyzed, including a novel CNN-based fusion technique, improving verification results to 2% EER and identification rank-1 results up to 99%.
- 4
- 4
- Dictionary-Based Methods for Object Recognition∗4
The role of algorithms based on SR and DL for object recognition are reviewed in this chapter.
- Sparse Representation-based Object Recognition4
This chapter outlines the Sparse Representation-based Classification (SRC) algorithm and presents its applications in robust biometrics recognition.
- Approximation Methods for the Recovery of Shapes and Images from Gradients4
This chapter reviews some approximation-based methods for surface reconstruction from the given nonintegrable gradient field with applications to 3D modeling and image reconstruction.
- Evaluation of ITS Technology for Bus Transit Systems4
Methods to forecast bus arrival times to a stop and the number passengers on board the bus and the ITS based strategies have the potential to improve connectivity between origins and destinations while reducing passenger waiting times are developed.
- 3
- Evaluating the generalizability of video-based assessment of intraoperative surgical skill in capsulorhexis3
Among the SSDA methods, Group Distributionally Robust Supervised Learning improved generalizability in most cases and Model performance improved with the domain adaptation methods the authors evaluated, but it fell short of within-dataset performance.
- Enhancing Knee MR Image Clarity through Image Domain Super-Resolution Reconstruction3
Although computationally intensive, this method offers a viable alternative to costly hardware upgrades and holds promise for improving diagnostic accuracy and generating more anatomically accurate models of the human body.
- 3
- SyncNoise: Geometrically Consistent Noise Prediction for Text-based 3D Scene Editing3
SyncNoise synchronously edits multiple views with 2D diffusion models while enforcing multi-view noise predictions to be geometrically consistent, which ensures global consistency in both semantic structure and low-frequency appearance.
- NBD-GAP: Non-Blind Image Deblurring without Clean Target Images3
This work views the non-blind image deblurring problem as a denoising problem, and presents a method that addresses these issues by performing Wiener filtering on a pair of blurry images with the corresponding blur kernels.
- Guest Editorial: Special Issue on “Computer Vision for All Seasons: Adverse Weather and Lighting Conditions”3
The last decade has witnessed tremendous progress in many sub-fields of computer vision, due to the increasing availability of large annotated datasets such as PASCAL VOC, but the performance is mainly benchmarked under good weather and favorable lighting conditions.
- A Joint Representation Learning and Feature Modeling Approach for One-class Recognition3
This work argues that both of these approaches have their own limitations; and a more effective solution can be obtained by combining the two, based on the combination of a generative framework and a one-class classification method.
- 3
- 3
- 3
- Active authentication using facial attributes3
The focus of this chapter is on face-based biometric authentication methods that first use the camera sensor images to detect the face of the users, next, they extract low-level features from the face images, and then apply their algorithm to the extracted features to authenticate the user.
- 3
- 3
- 3
- 3
- 3
- Remote Identification of Faces3
In this chapter, a remote face database which has been acquired in an unconstrained outdoor maritime environment is described and recognition performance of a subset of existing still image-based face recognition algorithms is discussed on the remote face data set.
- 3
- Image and Video-Based Biometrics3
This chapter discusses image and video-based biometric analysis involving face, iris and gait, and discusses several recent approaches to physiological biometrics based on Sparse Representations and Compressed Sensing.
- Multiscale directional filtering of noisy InSAR phase images3
This work presents a new approach for the problem of interferometric phase noise reduction in synthetic aperture radar interferometry based on the shearlet representation, and shows that this method performs significantly better than many competitive methods.
- 3
- 2
- MedCL: Learning Consistent Anatomy Distribution for Scribble-supervised Medical Image Segmentation2
This work proposes a scribble-supervised clustering-based framework, called MedCL, to learn the inherent anatomy distribution of medical labels, and evaluates the performance on three open datasets of regular structure, multiple organs and irregular pathology.
- Cinéaste: A Fine-grained Contextual Movie Question Answering Benchmark2
The analysis reveals that long-range temporal reasoning is a primary bottleneck, with the top open-source model achieving only 63.15\% accuracy, which underscores significant challenges in fine-grained contextual understanding and the need for advancements in long-form movie comprehension.
- Diffscaler: Enhancing the Generative Prowess of Diffusion Transformers2
This paper proposes DiffScaler, an efficient scaling strategy for diffusion models where a minimal amount of parameters are trained to adapt to different tasks, and finds that transformer-based diffusion models significantly outperform CNN-based diffusion models methods while performing fine-tuning over smaller datasets.
- I2I-Galip: Unsupervised Medical Image Translation Using Generative Adversarial CLIP2
A new image-to-image translation framework named Image-to-Image-Generative-Adversarial-CLIP (I2I-Galip) is proposed where a pre-trained multi-model foundation model (i.e., CLIP) is utilized to mitigate the need of separate generator-discriminator pairs for each source-target mapping while achieving better and more efficient multi-domain translation.
- 2
- Vision and Robotics2
The story of how some of the modern-day technologies the authors enjoy today trace their evolution from research sponsored by the Defense Advanced Research Projects Agency over the last 40 years is told.
- Quickest Multiple User Active Authentication2
The Multiple-user Quickest Intruder Detection (MQID) algorithm is formulated and the algorithm is extended to the data-efficient scenario where intruder detection is carried out with fewer observation samples.
- Prior-based Domain Adaptive Object Detection for Adverse Weather Conditions.2
This work proposes an unsupervised prior-based domain adversarial object detection framework for adapting the detectors to different weather conditions, and introduces a set of residual feature recovery blocks in the object detection pipeline to de-distort the feature space, resulting in further improvements.
- 2
- Robust sparse fourier transform based on the fourier projection-slice theorem2
This paper proposes a robust FPS-SFT (RFPS-S FT), which applies to noisy signals that contain off-grid frequencies; such signals arise in radar applications and the performance of the proposed method is demonstrated both theoretically and numerically.
- 2
- 2
- Recent Developments in Video-Based Face Recognition2
This chapter presents an overview of recent video-based face recognition methods, in particular, recent sparse coding-based, manifold- based, probabilistic, geometric model-based and dynamic model- based methods are reviewed.
- 2
- 2
- 2
- 2
- 2
- 2
- 1
- 1
- 1
- TransFIRA: Transfer Learning for Face Image Recognizability Assessment1
TransFIRA is introduced, a unified, geometry-driven framework for recognizability assessment that is encoder-specific, accurate, interpretable, and extensible across modalities, significantly advancing FIQA in accuracy, explainability, and scope.
- Low-Rank Adaptation-Based All-Weather Removal for Autonomous Navigation1
This paper proposes using Low-Rank Adaptation (LoRA) to efficiently adapt a pre-trained all-weather model to novel weather restoration tasks and introduces a LoRAbased fine-tuning method called LoRA-Align (LoRA-A) which seeks to align the singular vectors of the fine-tuned and pretrained weight matrices using Singular Value Decomposition (SVD).
- Generative Artificial Intelligence for Secure and Scalable Multimodal Eye Movement Datasets1
Generative AI is leveraged to create realistic videos of artificial eye movement, eliminating the need for real patient data, and these synthetic datasets have shown performance comparable to actual patient data in clinical tasks.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- T1: Scaling Diffusion Probabilistic Fields to High-Resolution on Unified Visual Modalities1
This work proposes a new model comprising of a view-wise sampling algorithm to focus on local structure learning, and incorporating additional guidance, e.g., text description, to complement the global geometry.
- 1
- 1
- 1
- Active Authentication on Mobile Devices1
This chapter provides a survey of recent face-based AA methods and proposes a system where users are continuously monitored after the initial access to the mobile device.
- 1
- Distance Preserving Projection Space of Symmetric Positive Definite Manifolds1
This paper proposes an optimized projection, based on building local and global sparse similarity graphs that encode the association of data points to the underlying subspace of each point in an optimized distance preserving projection space (DPS), which can be followed by any Euclidean-based classification algorithm.
- Energy Efficient Ad-hoc Routing Protocol for Disaster Scenario1
This paper has proposed a new technique based on link stability and energy aware, which is applied to Ad-hoc On Demand Distance Vector (AODV) protocol and it is observed that this proposed technique on AODV protocol consume less energy and gives better performance than different variant of A ODV.
- Compressive Sensing for Vision1
An overview of some of the recent works in computer vision and image understanding that make the use of compressive sampling and sparse representation and how sparse representation framework can lead to better reconstruction of images and surfaces from gradients is presented.
- 1
- 1
- Object Dependent Manifold Priors for Image Deconvolution1
A manifold based deconvolution algorithm by estimating a manifold from a set of natural images that exploits the availability of the sample class data for regularizing the deblurring problem is proposed.
- –
- –
- –
- –
- –
- From Imitation to Intuition: Intrinsic Reasoning for Open-Instance Video Classification–
DeepIntuit introduces an intrinsic reasoning framework that evolves open-instance video classification from imitation to intuition, followed by refinement using Group Relative Policy Optimization (GRPO) to enhance reasoning coherence through reinforcement learning.
- –
- Characterizing the robustness of multimodal AIS and imagery models for vessel image classification–
It is shown that the multi-modal imagery and AIS models can yield an up to a 13.3% increase in the average precision to identify a ship in imagery when 67% of pixels have been dropped from an image, confirming that the additional AIS modality can improve model robustness.
- –
- –
- Morphing Through Time: Diffusion-Based Bridging of Temporal Gaps for Robust Alignment in Change Detection–
This work presents a lightweight, modular pipeline that strengthens robustness without retraining the underlying change detection models, and significantly improves both registration accuracy and change detection performance, especially in scenarios with substantial spatial and temporal variations.
- –
- –
- –
- –
- Training Free Stylized Abstraction–
This work proposes a training-free framework that generates stylized abstractions from a single image using inference-time scaling in vision-language models (VLLMs) to extract identity-relevant features, and a novel cross-domain rectified flow inversion strategy that reconstructs structure based on style-dependent priors.
- –
- –
- –
- –
- –
- AttentiveGRUAE: An Attention-Based GRU Autoencoder for Temporal Clustering and Behavioral Characterization of Depression from Wearable Data–
AttentiveGRUAE, a novel attention-based gated recurrent unit (GRU) autoencoder designed for temporal clustering and prediction of outcome from longitudinal wearable data, demonstrates superior performance over baseline clustering, domain-aligned self-supervised, and ablated models in both clustering quality and stability.
- –
- –
- Accelerated MRI Reconstruction Using a Lightweight Recurrent Transformer: ReconFormer–
A recurrent transformer model, namely, ReconFormer, is proposed for MRI reconstruction which can iteratively reconstruct high fertility magnetic resonance images from highly under-sampled k-space data and achieves significant improvements over the state-of-the-art methods with better parameter efficiency.
- –
- Systematic Investigation of Selected Bio-molecules Potentially Effective in Inhibiting SARS-CoV-2/COVID-19 via In-Silico Analysis–
Several bioactive markers, including Nobiletin, Tangeretin, Epicatechin, Epigallocatechin gallate (EGCG), Ombuin, Tamarixetin, and Nimbolide, exhibited similar binding sites to synthetic drugs, emphasizing the importance of vaccination as a crucial measure to curb the spread of infection.
- Artificial Intelligence, Deep Learning, and Machine Learning Applications in Total Knee Arthroplasty–
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Discriminative Sparse Representations–
The role of algorithms based on SR and DL for object recognition are reviewed, in particular, supervised, unsupervised, weakly supervised, nonlinear kernel-based, convolutional sparse coding and analysis DL methods are reviewed.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Multicomposite wavelet estimation–
This paper shows how to take advantage of different wavelets with composite dilations to sparsely represent important features such as edges and texture independently, and apply these techniques to derive improved algorithms for image denoising.
- Sparse and redundant representations for inverse problems and recognition–
This research investigates the combination of domain adaptation, dictionary learning, object recognition, activity recognition, and shape representation in machine learning to solve the challenge of sparse representation in signal/Image processing.
- –
- –
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 2026-10-15. 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.