Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Pierre‐Marc JodoinSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works117 from public data
- 880
- Interactive deep learning method for segmenting moving objects334
The first machine learning method for ground truthing a video is proposed, based on a multi-resolution convolutional neural network with a cascaded architecture, for segmenting foreground moving objects pictured in surveillance videos.
- 200
- A Self-Training Approach for Point-Supervised Object Detection and Counting in Crowds123
A novel self-training approach named Crowd-SDNet that enables a typical object detector trained only with point-level annotations to estimate both the center points and sizes of crowded objects and proposes a confidence and order-aware refinement scheme to continuously refine the initial pseudo object sizes.
- 113
- Rethinking and Designing a High-Performing Automatic License Plate Recognition Approach99
The proposed VSNet outperforms state-of-the-art methods by more than 50% relative improvement on error rate, achieving >99% recognition accuracy on CCPD and AOLP datasets with 149 FPS inference speed and illustrates an outstanding generalization capability when evaluated on the unseen PKUData and CLPD datasets.
- A survey on occupancy perception for autonomous driving: The information fusion perspective84
This survey reviews the most recent works on 3D occupancy perception, and provides in-depth analyses of methodologies with various input modalities, and evaluates and analyze the occupancy perception performance of the state-of-the-art on the most popular datasets.
- 74
- TAFNet: A Three-Stream Adaptive Fusion Network for RGB-T Crowd Counting73
A three-stream adaptive fusion network named TAFNet, which uses paired RGB and thermal images for crowd counting, which achieves more than 20% improvement on mean average error and root mean squared error compared with state-of-the-art method.
- A Survey of Embodied Learning for Object-centric Robotic Manipulation63
A comprehensive survey of the latest advancements in embodied learning for object-centric robotic manipulation and categorize the existing work into three main branches: 1) Embodied perceptual learning, which aims to predict object pose and affordance through various data representations.
- 58
- Interactive Change-Aware Transformer Network for Remote Sensing Image Change Captioning50
An Interactive Change-Aware Transformer Network (ICT-Net) is proposed, able to extract and incorporate the most critical changes of interest in each encoder layer to improve change description generation and achieves a state-of-the-art performance.
- Open World Object Detection: A Survey47
This survey paper offers a thorough review of the OWOD domain, covering essential aspects, including problem definitions, benchmark datasets, source codes, evaluation metrics, and a comparative study of existing methods.
- Convolutional Neural Networks With Dynamic Regularization41
This article proposes a dynamic regularization method for CNNs that can dynamically adjust the regularization strength in the training procedure, thereby balancing the underfitting and overfitting of CNNs.
- Dense Point Prediction: A Simple Baseline for Crowd Counting and Localization33
A localization head is proposed to distinguish dense crowds trained by two loss functions, i.e., Negative-Suppressed Focal (NSF) loss and False-Positive (FP) loss, which balances the positive/negative examples and handles the false-positive predictions.
- 32
- Weakly-Supervised Part-Attention and Mentored Networks for Vehicle Re-Identification31
This paper proposes a weakly-supervised Part-Attention Network (PANet) and Part-Mentored Network (PMNet) for Vehicle Re-ID and adopts Homoscedastic Uncertainty to learn the optimal weighing of ID losses.
- 28
- 23
- 21
- NTIRE 2026 Challenge on Bitstream-Corrupted Video Restoration: Methods and Results19
The NTIRE 2026 Challenge on Bitstream-Corrupted Video Restoration provides a common benchmark for evaluating restoration methods under realistic corruption settings and provides useful insights for future research on robust video restoration under practical bitstream corruption.
- SGIFormer: Semantic-Guided and Geometric-Enhanced Interleaving Transformer for 3D Instance Segmentation19
This paper introduces a novel method, named SGIFormer, for 3D instance segmentation, which is composed of the Semantic-guided Mix Query (SMQ) initialization and the Geometric-enhanced Interleaving Transformer (GIT) decoder.
- SignEye: Traffic Sign Interpretation From Vehicle First-Person View18
A traffic guidance assistant (TGA) scenario application is developed to re-explore the role of traffic signs in ADS as a complement to popular autonomous technologies (such as obstacle perception) and results demonstrate that the TGA can provide complementary information to ADS beyond existing popular autonomous technologies.
- 18
- 18
- 18
- 16
- Bitstream-Corrupted JPEG Images are Restorable: Two-stage Compensation and Alignment Framework for Image Restoration15
A robust JPEG decoder is proposed, followed by a two-stage compensation and alignment framework to restore bitstream-corrupted JPEC images, and experimental results and ablation studies demonstrate the superiority of the proposed method.
- A Byte Sequence is Worth an Image: CNN for File Fragment Classification Using Bit Shift and n-Gram Embeddings14
This work proposes Byte2Image, a novel data augmentation technique, to introduce the neglected intra-byte information into file fragments and re-treat them as 2d gray-scale images, which allows us to capture both inter-byte and intra- byte correlations simultaneously through powerful convolutional neural networks (CNNs).
- End-to-end varifocal multiview images coding framework from data acquisition end to vision application end14
An end-to-end coding scheme for V FMV images is proposed, which provides a new paradigm for VFMV compression from data acquisition (source) end to vision application end, and extensive experiments demonstrate that the proposed coding scheme is superior to the comparison scheme in objective quality, subjective quality and computational complexity.
- HVG-3D: Bridging Real and Simulation Domains for 3D-Conditional Hand-Object Interaction Video Synthesis13
HVG-3D, a unified framework for 3D-aware hand-object interaction (HOI) video synthesis conditioned on explicit 3D representations, is proposed, which develops a diffusion-based architecture augmented with a 3D ControlNet, which encodes geometric and motion cues from 3D inputs to enable explicit 3D reasoning during video synthesis.
- 13
- Moving Towards Centers: Re-Ranking With Attention and Memory for Re-Identification13
This paper proposes a re-ranking network to predict the correlations between the probe and top-ranked neighbor samples, which surpasses the state-of-the-art re- ranking approaches on large-scale datasets by a significant margin.
- SSN: Stockwell Scattering Network for SAR Image Change Detection12
Stockwell scattering network (SSN) based on Stockwell transform (ST) is proposed, which provides noise-resilient feature representation and obtains state-of-the-art performance in SAR image change detection as well as high computational efficiency.
- REAL: Representation enhanced analytic learning for exemplar-free class-incremental learning11
A representation-enhanced analytic learning (REAL) is proposed, which achieves state-of-the-art performance on CIFAR-100, ImageNet-100 and ImageNet-1k benchmarks, outperforming exemplar-free methods and rivaling exemplar-based approaches.
- 11
- CaRe-Ego: Contact-aware relationship modeling for egocentric interactive hand-object segmentation11
A novel method called CaRe-Ego is proposed that achieves state-of-the-art performance by emphasizing the contact between hands and objects from two aspects and significantly outperforms existing methods with robust generalization capability.
- OccluTrack: Rethinking Awareness of Occlusion for Enhancing Multiple Pedestrian Tracking11
This work proposes an adaptive occlusion-aware multiple pedestrian tracker, OccluTrack, that outperforms state-of-the-art methods on MOTChallenge and DanceTrack datasets and develops a pose-guided re-identification module to extract discriminative part features for partially occluded pedestrians.
- 11
- PromptSR: Cascade Prompting for Lightweight Image Super-Resolution10
The experimental results demonstrate the superiority of the PromptSR method, which outperforms state-of-the-art lightweight SR methods in quantitative, qualitative, and complexity evaluations.
- 10
- 10
- 10
- ANNEXE: Unified Analyzing, Answering, and Pixel Grounding for Egocentric Interaction9
A unified ANNEXE model is designed to generate text- and pixel-level outputs utilizing multimodal large language models, which enables a comprehensive interpretation of egocentric interactions.
- 8
- 7
- Multi-Object Tracking with Tracked Object Bounding Box Association7
This paper proposes to incorporate a simple tracked object bounding box and overlapping prediction based on the current frame onto the CenterTrack algorithm and proposes an Intersection over Union (IOU) distance cost matrix in the association step instead of simple point displacement distance.
- Building Egocentric Procedural AI Assistant: Methods, Benchmarks, and Challenges6
This paper identifies three core tasks in EgoProceAssist: egocentric procedural error detection, egocentric procedural learning, and egocentric procedural question answering, and introduces two enabling dimensions: real-time and streaming video understanding, and proactive interaction in procedural contexts.
- Towards Blind Bitstream-corrupted Video Recovery: A Visual Foundation Model-driven Framework6
This paper proposes the first blind bitstream-corrupted video recovery framework that integrates visual foundation models with recovery model, which is adapted to different types of corruption and bitstream-level prompts and introduces a novel Corruption-aware Feature Completion (CFC) module.
- 6
- 6
- Interaction-aware Representation Modeling with Co-occurrence Consistency for Egocentric Hand-Object Parsing5
This work proposes an end-to-end Interaction-aware Transformer (InterFormer), which integrates three key components, i.e., a Dynamic Query Generator, a Dual-context Feature Selector (DFS), and the Conditional Co-occurrence (CoCo) loss, and achieves state-of-the-art performance on both the EgoHOS and the challenging out-of-distribution mini-HOI4D datasets.
- 5
- 3DGeoDet: General-Purpose Geometry-Aware Image-Based 3D Object Detection5
The predicted depth is utilized to learn voxel occupancy and optimize the voxelized 3D feature volume explicitly through the proposed voxel occupancy attention, and an implicit 3D representation is integrated with an implicit 3D representation, the truncated signed distance function (TSDF).
- LaSSM: Efficient Semantic-Spatial Query Decoding via Local Aggregation and State Space Models for 3D Instance Segmentation4
LaSSM is introduced, a hierarchical semantic-spatial query initializer to derive the query set from superpoints by considering both semantic cues and spatial distribution, achieving comprehensive scene coverage and accelerated convergence.
- Revealing the intrinsic ethical vulnerability of aligned large language models4
It is proved current alignment creates only local safety regions, leaving global adversarial paths intact, and empirically show widespread vulnerability across leading aligned models, showing a fundamental structural weakness in current aligned LLMs.
- EDVD-LLaMA: Explainable Deepfake Video Detection via Multimodal Large Language Model Reasoning4
This paper proposes the explainable deepfake video detection (EDVD) task and designs the EDVD-LLaMA multimodal, a large language model (MLLM) reasoning framework, which provides traceable reasoning processes alongside accurate detection results and trustworthy explanations.
- GaussianCross: Cross-modal Self-supervised 3D Representation Learning via Gaussian Splatting4
GaussianCross is presented, a novel cross-modal self-supervised 3D representation learning architecture integrating feed-forward 3D Gaussian Splatting techniques to address current challenges and shows a prominent parameter and data efficiency.
- Attack Anything: Blind DNNs via Universal Background Adversarial Attack4
The findings of this research substantiate the significant discrepancy between human and machine vision on the value of background variations, which play a far more critical role than previously recognized, necessitating a reevaluation of the robustness and reliability of DNNs.
- 4
- 4
- 4
- EVA02-AT: Egocentric Video-Language Understanding With Spatial-Temporal Rotary Positional Embeddings and Symmetric Optimization3
This paper introduces EVA02-AT, a suite of EVA02-based video-language foundation models tailored to egocentric video understanding tasks, and introduces the Symmetric Multi-Similarity (SMS) loss and a novel training framework that advances all soft labels for both positive and negative pairs, providing a more precise learning objective.
- From Semantics, Scene to Instance-awareness: Distilling Foundation Model for Open-vocabulary Grounded Situation Recognition3
This paper proposes Multimodal Interactive Prompt Distillation (MIPD), a novel framework that distills enriched multimodal knowledge from the foundation model, enabling the student Ov-GSR model to recognize unseen situations and be better aware of rare situations.
- PADetBench: Towards Benchmarking Physical Attacks against Object Detection3
Realistic simulation is explored utilizing realistic simulation to thoroughly and rigorously benchmark physical attacks with fairness under controlled physical dynamics and cross-domain transformation, resolving the problem of capturing identical adversarial images that cannot be achieved in the real world.
- 3
- 3
- EgoProceVQA: A Novel Egocentric Procedural Understanding Task with Self-Skill-Exploration Agent2
The Egocentric Procedural Understanding VQA task is introduced, which systematically evaluates egocentric procedural reasoning abilities of current MLLMs and agents through six types of key-step-centric questions, and EgoProceAgent, a self-skill-exploration agentic framework, a generic tool library for procedural understanding and a standardized sub-skill library shared across tools and models are designed.
- Future-Proofing Deepfake Detection by Integrating Audio, Video, and Text2
This perspective letter advocates for a new paradigm in deepfake detection that emphasizes the integration of audio, video, and textual content, and examines the limitations of current systems, including their over-reliance on outdated datasets and limited adversarial robustness.
- HAMMER: Harnessing MLLM via Cross-Modal Integration for Intention-Driven 3D Affordance Grounding2
This work proposes a novel framework that leverages emerging multimodal large language models (MLLMs) for interaction intention-driven 3D affordance grounding, namely HAMMER, and devise a hierarchical cross-modal integration mechanism to fully exploit the complementary information from the MLLM for 3D representation refinement.
- Accelerating Diffusion-based Video Editing via Heterogeneous Caching: Beyond Full Computing at Sampled Denoising Timestep2
HetCache is introduced, a training-free diffusion acceleration framework designed to exploit the inherent heterogeneity in diffusion-based masked video-to-video (MV2V) generation and editing, which reduces redundant attention operations while maintaining editing consistency and fidelity.
- 2
- 2
- Symmetric Multi-Similarity Loss for EPIC-KITCHENS-100 Multi-Instance Retrieval Challenge 20242
This report proposes a novel loss function, Symmetric Multi-Similarity Loss, which offers a more precise learning objective and achieves 63.76% average mAP and 74.25% average nDCG on the public leaderboard, demonstrating the effectiveness of the approach.
- 2
- 2
- 2
- Mixture-of-Modality-Experts with Holistic Token Learning for Fine-Grained Multimodal Visual Analytics in Driver Action Recognition1
The experimental results on the public benchmark show that the proposed MoME framework and the HTL strategy jointly outperform representative single-modal and multimodal baselines and improves subtle multimodal understanding and offers better interpretability.
- ProCal: Probability Calibration for Neighborhood-Guided Source-Free Domain Adaptation1
ProCal, a probability calibration method that dynamically calibrates neighborhood-based predictions through a dual-model collaborative prediction mechanism, is introduced, achieving a balance between knowledge retention and domain adaptation.
- EARL: Towards a Unified Analysis-Guided Reinforcement Learning Framework for Egocentric Interaction Reasoning and Pixel Grounding1
EARL is an Egocentric Analysis-guided Reinforcement Learning framework that explicitly transfers coarse interaction semantics to query-oriented answering and grounding, and adopts a two-stage parsing framework including coarse-grained interpretation and fine-grained response.
- GUI-C$^2$: Coarse-to-Fine GUI Grounding via Difficulty-Aware Reinforcement Learning1
This work proposes GUI-D, a data mining and difficulty scoring pipeline that identifies the training-worthy samples by proper testing and assigns difficulty scores to guide subsequent training weights, and employs an area-gated coarse-to-fine refinement mechanism that progressively narrows the visual field via model-internal uncertainty signals.
- From Semantics, Scene to Instance-awareness: Distilling Foundation Model for Grounded Open-vocabulary Situation Recognition1
This paper proposes Multimodal Interactive Prompt Distillation (MIPD), a novel framework that distills enriched multimodal knowledge from the foundation model, enabling the student Ov-GSR model to recognize unseen situations and be better aware of rare situations.
- 1
- 1
- 1
- 1
- Video sentence grounding with temporally global textual knowledge1
A Pseudo-query Intermediary Network (PIN) is proposed to achieve an improved alignment of visual and comprehensive pseudo-query features within the feature space through contrastive learning and enhances the feature alignment between visual and language for better temporal grounding.
- 1
- 1
- 1
- 1
- –
- –
- –
- –
- –
- Evolution-Inspired Sample Competition for Deep Neural Network Optimization–
Natural Selection (NS) is presented, a novel evolution-inspired optimization method that explicitly incorporates competitive interactions into deep network training and enables more adaptive and balanced model optimization.
- SP-MoMamba: Superpixel-driven Mixture of State Space Experts for Efficient Image Super-Resolution–
The proposed Superpixel-driven State Space Model (SP-SSM) organizes spatially coherent features into compact region-level tokens and performs global sequence modeling over these content-aware representations, reducing redundant computation while facilitating long-range structural interaction.
- –
- MASS: Mesh-inellipse Aligned Deformable Surfel Splatting for Hand Reconstruction and Rendering from Egocentric Monocular Video–
This work introduces the mesh-aligned Steiner Inellipse and fractal densification for mesh-to-surfel conversion that initiates high-resolution 2D Gaussian surfels from coarse parametric hand meshes, providing surface representation with photorealistic rendering potential.
- –
- –
- –
- Blind object detectors via transferable background adversarial attack–
The findings of this research substantiate the significant discrepancy between human and machine vision on the value of background variations, which play a far more critical role than previously recognized, necessitating a reevaluation of the robustness and reliability of DNNs.
- –
- Probabilistic Mixture of Hyperbolic Mamba for Few-Shot Class-Incremental Learning–
This paper presents a novel framework named Probabilistic Mixture of Hyperbolic State Space Experts (PmH-SSE), and proposes the probabilistic mixture of Mamba to increase the model's flexibility in handling non-stationary data streams in FSCIL and enhance the stability of high-parameter models in few-shot conditions.
- –
- HSNet: Heterogeneous Subgraph Network for Single Image Super-resolution–
The Heterogeneous Subgraph Network (HSNet) is proposed, a novel framework that efficiently leverages graph modeling while maintaining computational feasibility, and achieves state-of-the-art performance, effectively balancing reconstruction quality with computational efficiency.
- –
- –
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-13. 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.