Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks90 from public data
- 428
- Evo-ViT: Slow-Fast Token Evolution for Dynamic Vision Transformer317
Evo-ViT is presented, a self-motivated slow-fast token evolution approach for vision transformers that can accelerate vanilla transformers of both flat and deep-narrow structures from the very beginning of the training process.
- VITA: Towards Open-Source Interactive Omni Multimodal LLM204
VITA, the first-ever open-source Multimodal Large Language Model (MLLM) adept at simultaneous processing and analysis of Video, Image, Text, and Audio modalities, and meanwhile has an advanced multimodal interactive experience is introduced.
- Asymmetric Co-Teaching for Unsupervised Cross-Domain Person Re-Identification172
An asymmetric co-teaching framework, which resists noisy labels by cooperating two models to select data with possibly clean labels for each other, which can consistently benefit most clustering based methods, and boost the state-of-the-art adaptation accuracy.
- Toward an Expert Level of Lung Cancer Detection and Classification Using a Deep Convolutional Neural Network166
The three-dimensional convolutional neural network described in this article demonstrated both high sensitivity and high specificity in classifying pulmonary nodules regardless of diameters as well as superiority compared with manual assessment.
- 159
- High-Dimensional Dense Residual Convolutional Neural Network for Light Field Reconstruction153
This work formulate light field super-resolution (LFSR) as tensor restoration and develop a learning framework based on a two-stage restoration with 4-dimensional (4D) convolution, which allows the model to learn the features capturing the geometry information encoded in multiple adjacent views.
- Temporal Modulation Network for Controllable Space-Time Video Super-Resolution130
A Temporal Modulation Network (TMNet) to interpolate arbitrary intermediate frame(s) with accurate high-resolution reconstruction and demonstrates that the TMNet outperforms previous STVSR methods.
- Removing the Background by Adding the Background: Towards Background Robust Self-supervised Video Representation Learning115
Self-supervised learning has shown great potentials in improving the video representation ability of deep neural networks by getting supervision from the data itself, but some of the current methods tend to cheat from the background, so this work proposes to remove the background impact by adding the background.
- Viewpoint-Aware Loss with Angular Regularization for Person Re-Identification106
This work proposes a novel approach to viewpoint-aware adaptive label smoothing regularization (VALSR) that assigns the adaptive soft label to feature representation and demonstrates that it outperforms the state-of-the-art supervised Re-ID methods.
- 89
- 81
- Cantor: Inspiring Multimodal Chain-of-Thought of MLLM74
It is argued that converging visual context acquisition and logical reasoning is pivotal for tackling visual reasoning tasks and an innovative multimodal CoT framework, termed Cantor, characterized by a perception-decision architecture is proposed, demonstrating the efficacy of the proposed framework.
- Enhancing Unsupervised Video Representation Learning by Decoupling the Scene and the Motion73
This work proposes to decouple the scene and the motion (DSM) with two simple operations, so that the model attention towards the motion information is better paid and the impact of the scene is weakened while the temporal sensitivity of the network is further enhanced.
- 72
- Mining Approximate Frequent Itemsets In the Presence of Noise: Algorithm and Analysis70
This paper developed and implemented an algorithm to mine AFIs that generalizes the level-wise enumeration of frequent itemsets by allowing noise, and proposes the noise-tolerant support threshold, a relaxed version of support, which varies with the length of the itemset and the noise threshold.
- Training-free Transformer Architecture Search68
Experimental results demonstrate that the TF- TAS achieves a competitive performance against the state-of-the-art manually or automatically design ViT architectures, and it promotes the searching efficiency in ViT search space greatly.
- Do Not Disturb Me: Person Re-identification Under the Interference of Other Pedestrians63
A novel deep network termed Pedestrian-Interference Suppression Network (PISNet) is presented, which leverages a Query-Guided Attention Block (QGAB) to enhance the feature of the target in the gallery, under the guidance of the query.
- Co-Salient Object Detection with Co-Representation Purification57
This paper proposes a Co-Representation Purification Purification (CoRP) method aiming at searching noise-free co-representation, and uses the prediction to iteratively reduce irrelevant embeddings in the co- representation.
- 57
- Training-Free Group Relative Policy Optimization53
This work proposes Training-Free Group Relative Policy Optimization (Training-Free GRPO), a cost-effective solution that enhances LLM agent performance without any parameter updates and demonstrates that Training-Free GRPO, when applied to DeepSeek-V3.1-Terminus, significantly improves out-of-domain performance.
- NOH-NMS51
This work proposes \heatmapname (\heatmapnameshort ), which pinpoints the objects nearby each proposal with a Gaussian distribution, together with \nmsname, which dynamically eases the suppression for the space that might contain other objects with a high likelihood.
- 49
- 46
- VITA-Audio: Fast Interleaved Cross-Modal Token Generation for Efficient Large Speech-Language Model42
VITA-Audio is the first multi-modal large language model capable of generating audio output during the first forward pass, enabling real-time conversational capabilities with minimal latency and significantly outperforms open-source models of similar model size on multiple benchmarks.
- SinKD: Sinkhorn Distance Minimization for Knowledge Distillation41
The Sinkhorn Knowledge Distillation (SinKD) is proposed that exploits the Sinkhorn distance to ensure a nuanced and precise assessment of the disparity between teacher and student distributions and gets rid of sample-wise KD that restricts the perception of divergence in each teacher-student sample pair.
- Filter Grafting for Deep Neural Networks36
An entropy-based criterion to measure the information of filters and an adaptive weighting strategy for balancing the grafted information among networks are developed and performed to improve the representation capability of Deep Neural Networks.
- 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.
- 34
- Turning a CLIP Model Into a Scene Text Spotter27
This work exploits the potential of the large-scale Contrastive Language-Image Pretraining model to enhance scene text detection and spotting tasks, transforming it into a robust backbone, FastTCM-CR50, which showcases robust few-shot training capabilities.
- D3G: Exploring Gaussian Prior for Temporal Sentence Grounding with Glance Annotation26
This study investigates a recently proposed glance-supervised temporal sentence grounding task, which requires only single frame annotation (referred to as glance annotation) for each query and proposes a Dynamic Gaussian prior based Grounding framework with Glance annotation, which outperforms the state-of-the-art weakly supervised methods by a large margin and narrows the performance gap compared to fully supervised methods.
- 26
- Dig into Multi-modal Cues for Video Retrieval with Hierarchical Alignment25
This paper leverages the hierarchical video-text alignment to fully explore the detailed diverse characteristics in multi-modal cues for fine-grained alignment with local semantics from phrases, as well as to capture a high-level semantic correspondence.
- A Coarse-to-fine Pyramidal Model for Person Re-identification via Multi-Loss Dynamic Training.24
A novel coarse-to-fine pyramid model is proposed that not only incorporates local and global information, but also integrates the gradual cues between them that is able to match the cues at different scales and then search for the correct image of the same identity even when the image pair are not aligned.
- Learning Canonical View Representation for 3D Shape Recognition with Arbitrary Views22
A canonical view representation is proposed to tackle the challenge of recognizing 3D shapes from arbitrary views, i.e., arbitrary numbers and positions of viewpoints, and achieves competitive results under the fixed viewpoint settings and significantly outperforms the applicable methods under the arbitrary view setting.
- Rethinking Temporal Fusion for Video-Based Person Re-Identification on Semantic and Time Aspect22
A novel general temporal fusion framework to aggregate frame features on both semantic aspect and time aspect and a multi-stage fusion network is explored to fuse richer frame features at multiple semantic levels, which can effectively reduce the information loss caused by the traditional single- stage fusion.
- Ask&Confirm: Active Detail Enriching for Cross-Modal Retrieval with Partial Query21
This work introduces the partial-query problem and extensively analyze its influence on text-based image retrieval, and proposes a novel retrieval framework that conducts the interactive process in an Ask-and-Confirm fashion, where AI actively searches for discriminative details missing in the current query, and users only need to confirm AI’s proposal.
- One for More: Selecting Generalizable Samples for Generalizable ReID Model20
A one-for-more training objective that directly takes the generalization ability of selected samples as a loss function and learn a sampler to automatically select generalizable samples that can effectively improve the ReID model training and boost the performance of ReID models.
- SAMSVM: A tool for misalignment filtration of SAM-format sequences with support vector machine20
Application of SAMSVM to actual sequencing data resulted in filtration of misaligned reads and correction of variant calling, indicating that the model built usingSAMSVM was accurate in misalignment detection.
- 19
- 19
- Reciprocal normalization for domain adaptation17
Reciprocal Normalization is proposed, a novel normalization method that outperforms existing normalization counterparts by a large margin and helps state-of-the-art adaptation approaches achieve better results.
- PR-Net: Preference Reasoning for Personalized Video Highlight Detection16
A simple yet efficient preference reasoning framework (PR-Net) to explicitly take the diverse interests into account for frame-level highlight prediction, which significantly outperforms state-of-the-art methods with a relative improvement of 12% in mean accuracy precision.
- 16
- Image generation and constrained two-stage feature fusion for person re-identification15
An improved generator and a constrained two-stage fusion network are proposed and a novel gesture discriminator embedded into the generator is used to calculate the completeness of skeleton pose images.
- Learning fused features with parallel training for person re-identification13
An efficient joint SIR and CIR learning strategy – cross-image features fusion strategy (CFFS), which fuses cross- image information through parallel training is proposed, which can achieve favorable accuracy while compared with state-of-the-arts.
- 13
- MMICT: Boosting Multi-Modal Fine-Tuning with In-Context Examples12
Multi-Modal In-Context Tuning (MMICT) is introduced, a novel multi-modal fine-tuning paradigm that boosts multi-modal fine-tuning by fully leveraging the promising ICL capability of Multi-Modal LLMs (MM-LLMs).
- FIPO: Free-form Instruction-oriented Prompt Optimization with Preference Dataset and Modular Fine-tuning Schema10
The FIPO framework uses a modular APO template that dynamically integrate the naive task instruction, optional instruction responses, and optional ground truth to produce finely optimized prompts, and is named Free-from Instruction-oriented Prompt Optimization (FIPO).
- 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
- Coarse-to-Fine: Learning Compact Discriminative Representation for Single-Stage Image Retrieval10
This work designs a Coarse-to-Fine framework to learn Compact Discriminative representation (CFCD) for end-to-end single-stage image retrieval-requiring only image-level labels and proposes a mechanism which attentively selects prominent local descriptors and infuse fine-grained semantic relations into the global representation by a hard negative sampling strategy to optimize inter-class distinctiveness at a global scale.
- 10
- 10
- Computational Light Field Generation Using Deep Nonparametric Bayesian Learning10
This paper presents a deep nonparametric Bayesian method to synthesize a light field from a single image and shows that this method outperforms several state-of-the-art novel view synthesis methods in delivering good image resolution.
- 9
- 9
- 8
- Multimodal Inplace Prompt Tuning for Open-set Object Detection6
This study introduces a redundancy assessment metric to identify uniform attention patterns and incorporates multimodal inplace prompt tuning (MIPT) to enrich the text prompt with visual clues in areas with high redundancy.
- 5
- 5
- FactGuard: Leveraging Multi-Agent Systems to Generate Answerable and Unanswerable Questions for Enhanced Long-Context LLM Extraction4
By implementing efficient data selection and generation within the multi-agent collaborative framework, this method significantly reduces the traditionally high costs associated with manual annotation and provides valuable insights for the training and optimization of LLMs.
- Multi-dataset Detection with Transformers4
A multi-dataset detector using the transformer (MDT) to enhance the effectiveness of the fusion of multiple datasets, and proposes alternative learning to suppress the noisy data to speed up the training of big data.
- 4
- 4
- 4
- Devil's in the Detail: Graph-based Key-point Alignment and Embedding for Person Re-ID4
An end-to-end ReID method that integrates both key recognition patterns into a unified framework, called Siamese-GCN, and achieves state-of-the-art performance on three public datasets is proposed.
- 4
- Towards Robust Text Retrieval with Progressive Learning3
Extensive experiments conducted on C-MTEB and DuReader demonstrate that PEG surpasses state-of-the-art embeddings in retrieving true positives, highlighting its significant potential for applications in LLMs.
- PAC-Net: Highlight Your Video via History Preference Modeling3
This work designs a Decision Boundary Customizer module to dynamically generate the user-adaptive highlight classifier from the preference-related user history, and introduces Mini-History (Mi-Hi) mechanism to capture more fine-grained user-specific preferences.
- 3
- 2
- DREAM: Document Reconstruction via End-to-end Autoregressive Model2
An innovative autoregressive model specifically designed for document reconstruction, referred to as Document Reconstruction via End-to-end Autoregressive Model (DREAM), which transmutes the text image into a sequence of document reconstruction in a comprehensive, end-to-end process, encapsulating a broader spectrum of document element information.
- Leveraging Open Knowledge for Advancing Task Expertise in Large Language Models2
An efficient and scalable pipeline to cost-efficiently produce task experts where K-shot data intervene in selecting the most promising expert candidates and the task-relevant instructions and the insistence on diversity is developed.
- 2
- 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.
- RESTORE: Towards Feature Shift for Vision-Language Prompt Learning1
To prevent feature misalignment, a feature shift consistency is introduced to synchronize inter-modal feature shifts by measuring and regularizing the magnitude of discrepancy during prompt tuning, and a "surgery"block is proposed to avoid short-cut hacking.
- 1
- Research on Crop Growth Period Estimation Based on Fusion Features1
A crop growth period estimation method based on fusion features is proposed, where the HOG features, SILTP features and color features are fused and XQDA is used to measure the similarity to classify and identify the growing period of crops.
- Synthesizing Virtual-Real Artworks Using Sun Orientation Estimation.1
This work proposes an end to end way of outdoor sun orientation estimation via a novel deep convolutional neural network (DCNN), which directly outputs the orientation of the sun from an outdoor image, which outperforms model state-of-the-art DCNN based methods.
- 1
- 1
- –
- –
- –
- –
- –
- DGD: Densifying the Knowledge of Neural Networks with Filter Grafting and Knowledge Distillation.–
The impacts of the two techniques are surprisingly complementary: distillation mostly enhances the knowledge of valid filters while grafting mostly reactivates invalid filters, helping to design a unified training framework called DGD, where distillation and grafting are naturally combined to increase the knowledge density inside the filters given a fixed model structure.
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.