Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks87 from public data
- Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysis1,687
Video-MME is introduced, the first-ever full-spectrum, Multi-Modal Evaluation benchmark of MLLMs in Video analysis, and it is revealed that Gemini 1.5 Pro is the best-performing commercial model, significantly outperforming the open-source models with an average accuracy of 75%, compared to 71.9% for GPT-4o.
- Contrastive Learning for Compact Single Image Dehazing926
This paper proposes a novel contrastive regularization (CR) built upon contrastive learning to exploit both the information of hazy images and clear images as negative and positive samples, respectively and develops a compact dehazing network based on autoencoder-like framework.
- Towards Optimal Structured CNN Pruning via Generative Adversarial Learning585
This paper proposes an effective structured pruning approach that jointly prunes filters as well as other structures in an end-to-end manner and effectively solves the optimization problem by generative adversarial learning (GAL), which learns a sparse soft mask in a label-free and an end to end manner.
- Accelerating Convolutional Networks via Global & Dynamic Filter Pruning274
This paper proposes a novel global & dynamic pruning (GDP) scheme to prune redundant filters for CNN acceleration that achieves superior performance to accelerate several cutting-edge CNNs on the ILSVRC 2012 benchmark.
- Fusion-Mamba for Cross-Modality Object Detection190
This work investigates a novel cross-modality fusion approach by associating cross-modal features in a hidden state space based on an improved Mamba with a gating attention mechanism, and establishes a new baseline for cross-modality object detection.
- 180
- Farewell to Mutual Information: Variational Distillation for Cross-Modal Person Re-Identification172
A new strategy, Variational Self-Distillation (VSD), is presented, which provides a scalable, flexible and analytic solution to essentially fitting the mutual information but without explicitly estimating it, and highlights the need to rethink the way to estimate mutual information.
- 164
- Exploiting Kernel Sparsity and Entropy for Interpretable CNN Compression142
KSE is capable of simultaneously compressing each layer in an efficient way, which is significantly faster compared to previous data-driven feature map pruning methods, and significantly outperforms state-of-the-art methods.
- 106
- Towards Compact CNNs via Collaborative Compression105
A Collaborative Compression scheme, which joints channel pruning and tensor decomposition to compress CNN models by simultaneously learning the model sparsity and low-rankness and proposes multi-step heuristic compression to remove redundant compression units step-by-step.
- PAMS: Quantized Super-Resolution via Parameterized Max Scale104
A new quantization scheme termed PArameterized Max Scale (PAMS), which applies the trainable truncated parameter to explore the upper bound of the quantization range adaptively and a structured knowledge transfer (SKT) loss is introduced to fine-tune the quantized network.
- Aligning and Prompting Everything All at Once for Universal Visual Perception100
APE advances the convergence of detection and grounding by reformulating language-guided grounding as open-vocabulary detection, which efficiently scales up model prompting to thousands of category vocab-ularies and region descriptions while maintaining the ef-fectiveness of cross-modality fusion.
- 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.
- AttriCLIP: A Non-Incremental Learner for Incremental Knowledge Learning78
This paper proposes a non-incremental learner, named AttriCLIP, to incrementally extract knowledge of new classes or tasks, built upon the pre-trained visual-language model CLIP and compares it with CLIP-based and previous state-of-the-art continual learning methods in realistic settings with domain-shift and long-sequence learning.
- Comprehensive Regularization in a Bi-directional Predictive Network for Video Anomaly Detection78
A novel bi-directional architecture with three consistency constraints to comprehensively regularize the prediction task from pixel-wise, cross-modal, and temporal-sequence levels to outperforms advanced anomaly detectors and achieves state-of-the-art results.
- A Challenger to GPT-4V? Early Explorations of Gemini in Visual Expertise68
A preliminary exploration of Gemini Pro's visual understanding proficiency, which comprehensively covers four domains: fundamental perception, advanced cognition, challenging vision tasks, and various expert capacities, demonstrates the potential of Gemini to be a strong challenger to GPT-4V.
- MISSU: 3D Medical Image Segmentation via Self-Distilling TransUNet64
This work proposes to self-distill a Transformer-based UNet for medical image segmentation, which simultaneously learns global semantic information and local spatial-detailed features and achieves the best performance over previous state-of-the-art methods.
- Towards Compact Single Image Super-Resolution via Contrastive Self-distillation61
A novel contrastive self-distillation (CSD) framework to simultaneously compress and accelerate various off-the-shelf SR models to improve the quality of SR images and PSNR/SSIM via explicit knowledge transfer is proposed.
- Masked face detection via a modified LeNet54
A modified LeNet, termed MLeNet, is presented, which modifies the number of units in output layer of LeNet to suit for a specific classification, which reduces overfitting and improves the performance with a small quantity of training samples.
- Towards convolutional neural networks compression via global error reconstruction47
A global error reconstruction method termed GER is presented, which firstly leverages an SVD-based low-rank approximation to coarsely compress the parameters in the fully connected layers in a layerwise manner, and jointly optimized in a global perspective via back-propagation.
- 46
- WaveMamba: Wavelet-Driven Mamba Fusion for RGB-Infrared Object Detection45
WaveMamba is proposed, a cross-modality fusion method that efficiently integrates the unique and complementary frequency features of RGB and IR decomposed by Discrete Wavelet Transform (DWT).
- 38
- DisCo: Remedying Self-supervised Learning on Lightweight Models with Distilled Contrastive Learning34
It is found that the final inherent embedding of the mainstream SSL methods contains the most important information, and it is proposed to distill the final embedding to maximally transmit a teacher’s knowledge to a lightweight model by constraining the lastEmbedding of the student to be consistent with that of the teacher.
- 24
- Pruning Blocks for CNN Compression and Acceleration via Online Ensemble Distillation24
This paper proposes an online ensemble distillation method to automatically prune blocks/layers of a target network by transferring the knowledge from a strong teacher in an end-to-end manner, and employs the fast iterative shrinkage-thresholding algorithm to fast and reliably remove the redundant blocks.
- ESPACE: Accelerating Convolutional Neural Networks via Eliminating Spatial and Channel Redundancy24
The first attempt of reducing spatial and channel redundancy directly from the visual input for CNNs acceleration is made, which demonstrates new state-of-the-art acceleration performance by a factor of 5.48* and 4.12* speedup on AlexNet and GoogLeNet, respectively.
- Adaptive Hierarchy-Branch Fusion for Online Knowledge Distillation22
A novel Adaptive Hierarchy-Branch Fusion framework for Online Knowledge Distillation, termed AHBF-OKD, is proposed, which designs hierarchical branches and adaptive hierarchy-branch fusion module to boost the model diversity and aggregate complementary knowledge.
- 21
- Novelty Detection via Contrastive Learning with Negative Data Augmentation20
This work introduces a novel decoder-encoder framework that has significant superiority over cutting-edge novelty detectors and achieves new state-of-the-art results on various novelty detection benchmarks, e.g. CIFAR10 and DCASE.
- Neural Network Compression via Learnable Wavelet Transforms19
This paper shows how the fast wavelet transform can be used to compress linear layers in neural networks and has significantly fewer parameters yet still perform competitively with the state-of-the-art on synthetic and real-world RNN benchmarks.
- TimeSoccer: An End-to-End Multimodal Large Language Model for Soccer Commentary Generation18
To support long video understanding of soccer matches, MoFA-Select is introduced, a training-free, motion-aware frame compression module that adaptively selects representative frames via a coarse-to-fine strategy, and incorporates complementary training paradigms to strengthen the model's ability to handle long temporal sequences.
- 18
- AQ-DETR: Low-Bit Quantized Detection Transformer with Auxiliary Queries16
A new quantization strategy based on Auxiliary Queries for DETR (AQ-DETR), aiming to enhance the capacity of quantized queries, and a layer-by-layer distillation is proposed to reduce the quantization error between quantized attention and full-precision counterpart.
- Dietary Habits Modify the Association of Physical Exercise with Cognitive Impairment in Community-Dwelling Older Adults15
Physical exercise was not found to be significantly protective with respect to cognitive impairment in older adults unless they had specific dietary habits, and dietary habits should be emphasized when investigating the beneficial effects of physical exercise on cognitive function in community-dwelling older adults.
- 14
- 12
- 12
- Prior Knowledge-driven Dynamic Scene Graph Generation with Causal Inference12
This work formulated into two sub-tasks: video-specific commonsense graph generation (VsCG) and causal inference (CI), which aims to alleviate spurious correlation problems and achieves state-of-the-art performance.
- Learn from Concepts: Towards the Purified Memory for Few-shot Learning12
A novel purified memory mechanism is presented that simulates the recognition process of human beings and enables the model to purify the information from semantic labels and progressively learn consistent, stable, and expressive concepts when episodes are trained one by one.
- LIPT: Latency-Aware Image Processing Transformer10
This paper proposes a novel non-volatile sparse masking self-attention that utilizes a pre-computing sparse mask to capture contextual information from a larger window with no extra computation overload, and devise the low-latency proportion LIPT block that substitutes memory-intensive operators with the combination of self-attention and convolutions to achieve practical speedup.
- IDMR: Towards Instance-Driven Precise Visual Correspondence in Multimodal Retrieval10
Instance-Driven Multimodal Image Retrieval (IDMR), a novel task requiring models to retrieve images containing the same instance as a query image while matching a text-described scenario, is designed and experimental results demonstrate previous models' limitations in instance-aware retrieval and highlight the potential of MLLM for advanced retrieval applications.
- A General and Efficient Training for Transformer via Token Expansion10
This paper proposes a novel token growth scheme Token Expansion (termed ToE) to achieve consistent training acceleration for ViTs, and introduces an “initialization-expansion-merging” pipeline to maintain the integrity of the intermediate feature distribution of original transformers, preventing the loss of crucial learnable information in the training process.
- 10
- Knowledge Distillation with Multi-granularity Mixture of Priors for Image Super-Resolution10
MiPKD, a multi-granularity mixture of prior KD framework is presented to facilitate efficient SR model through the feature mixture in a unified latent space and stochastic network block mixture to demonstrate the effectiveness of the proposed MiPKD method.
- 10
- Nutritional risk and adverse health outcomes in Chinese community-dwelling older adults: A study based on the Elderly Nutritional Indicators for Geriatric Malnutrition Assessment (ENIGMA)9
ENIGMA could serve as a nutritional risk screening tool that has a robust role in predicting cognitive impairment, functional limitation, and frailty in Chinese community-dwelling older adults and has the potential to guide early nutritional intervention in communities and primary care settings in China.
- THANet: Transferring Human Pose Estimation to Animal Pose Estimation9
The proposed vision transformer and cross-domain method significantly improve the model’s accuracy and generalization ability for APE, and is inspired by the success of ViTPose in HPE.
- Interpretable Neural Network Decoupling9
A novel architecture decoupling method to interpret the network from a perspective of investigating its calculation paths, and introduces a novel architecture controlling module in each layer to encode the network architecture by a vector.
- Dietary Habits and Depression in Community-Dwelling Chinese Older Adults: Cross-Sectional Analysis of the Moderating Role of Physical Exercise8
Physical exercise may conceal the potential protective effects of some healthy dietary habits in terms of depression and counteract the detrimental effects of the unhealthy habits when physical exercise cannot be performed.
- 8
- 8
- Explicit Invariant Feature Induced Cross-Domain Crowd Counting8
The main idea is to explicitly extract domain-invariant features from both source and target domains, which builds a bridge to transfer more rich knowledge between two domains.
- 7
- 7
- Image coding using wavelet transform and classified vector quantisation6
A new image coding scheme using the wavelet transform and classified vector quantisation is presented and results show that the proposed scheme has a better performance than those utilising current scalar orvector quantisation schemes.
- 5
- 5
- DCS-RISR: Dynamic Channel Splitting for Efficient Real-world Image Super-Resolution5
This paper proposes a novel Dynamic Channel Splitting scheme for efficient Real-world Image Super-Resolution, termed DCS-RISR, which not only achieves the best trade-off between computation/parameter and PSNR/SSIM metric, but also effectively handles real-world images with different degradation levels.
- FZUImageReg: A toolbox for medical image registration and dose fusion in cervical cancer radiotherapy5
A toolbox (FZUImageReg), which is a user-friendly dose fusion system based on hybrid image registration for radiation dose assessment in cervical cancer radiotherapy, and can achieve accurate registration between EBRT and HDR-BT images and provide radiation dose warping and fusion results in terms of high accuracy and efficiency.
- Kumaraswamy Wavelet for Heterophilic Scene Graph Generation4
This work proves rigorously that when the heterophily on the scene graph increases, the spectral energy gradually shifts towards the high-frequency part and proposes the Kumaraswamy Wavelet Graph Neural Network (KWGNN), which leverages complementary multi-group Kumaraswamy wavelets to cover all frequency bands.
- 4
- Altered auditory processes pattern predicts cognitive decline in older adults: different modalities with aging4
The altered MMN model exists in different aging stages and it’s a promising electrophysiological predictor of cognitive decline in older adults.
- Filter Pruning for Efficient CNNs via Knowledge-driven Differential Filter Sampler4
A novel Knowledge-driven Differential Filter Sampler~(KDFS) with Masked Filter Modeling~(MFM) framework for filter pruning, which globally prunes the redundant filters based on the prior knowledge of a pre-trained model in a differential and non-alternative optimization.
- Dynamic Neural Network Decoupling.4
A novel architecture decoupling method is proposed, which dynamically discovers the hierarchical path consisting of activated filters for each input image, and subsequently disentangles the filters by limiting the outputs of filter during training.
- 3
- 2
- Wandering and feeling the Scenes: Body-Aware Diffusion for 3D Human Motion Generation2
TheBA-IDM is proposed, which enables fine-grained control of human motion within a scene by leveraging multimodal information, and is designed to operate on de-identified motion data, effectively protecting user privacy throughout the process.
- 2
- 2
- Low complexity sub-band image coding with pseudo QMF and pyramidal lattice VQ2
The authors present a codec which combines the efficiency and effectiveness of both sub-band coding and lattice quantisation, and the proposed method was implemented and the computer simulation results are presented.
- 1
- Towards Universal Perception through Language-Guided Open-World Object Detection1
UniPerception is proposed, a novel universal perception framework based on open-vocabulary object detection that achieves strong open-world generalization and multi-modal understanding, outperforming the existing state-of-the-art methods and establishing itself as a unified, instruction-driven perception system.
- Generation Properties of Stochastic Interpolation under Finite Training Set1
The theoretical analysis reveals that, in the presence of estimation errors, the stochastic generation process effectively produces convex combinations of training samples corrupted by a mixture of uniform and Gaussian noise.
- Sex-Specific Association of Body Mass Index with Hippocampal Subfield Volume and Cognitive Function in Non-Demented Chinese Older Adults1
Sex-specific associations between BMI and hippocampal subfield volumes and cognitive performance are revealed, providing valuable insights into the development of effective interventions for the early prevention of cognitive decline.
- HUWSOD: Holistic Self-training for Unified Weakly Supervised Object Detection1
This paper introduces a unified, high-capacity weakly supervised object detection network called HUWSOD, which utilizes a comprehensive self-training framework without needing external modules or additional supervision, and indicates that randomly initialized boxes, although significantly different from well-designed offline object proposals, are effective for WSOD training.
- 1
- 1
- Towards deep neural network compression via learnable wavelet transforms1
This paper shows how the fast wavelet transform can be used to compress linear layers in neural networks and has significantly fewer parameters yet still perform competitively with the state-of-the-art on synthetic and real-world RNN benchmarks.
- 1
- CLIP-Map: Structured Matrix Mapping for Parameter-Efficient CLIP Compression–
This paper proposes a novel mapping-based CLIP compression framework, CLIP-Map, which leverages learnable matrices to map and combine pretrained weights by Full-Mapping with Kronecker Factorization, aiming to preserve as much information from the original weights as possible.
- –
- –
- –
- –
- –
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.