Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks71 from public data
- GCNet: Non-Local Networks Meet Squeeze-Excitation Networks and Beyond2,184
A simplified network based on a query-independent formulation, which maintains the accuracy of NLNet but with significantly less computation is created, and this simplified design shares similar structure with Squeeze-Excitation Network (SENet), which generally outperforms both simplified NLNet and SENet on major benchmarks for various recognition tasks.
- End-to-End Semi-Supervised Object Detection with Soft Teacher698
This paper presents an end-to-end semi-supervised object detection approach, in contrast to previous more complex multi-stage methods, and proposes two simple yet effective techniques within this framework: a soft teacher mechanism where the classification loss of each unlabeled bounding box is weighed by the classification score produced by the teacher network.
- EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty686
This paper introduces EAGLE (Extrapolation Algorithm for Greater Language-model Efficiency), a simple yet highly efficient speculative sampling framework that effectively resolves the uncertainty, enabling precise second-to-top-layer feature prediction with minimal overhead.
- Side Adapter Network for Open-Vocabulary Semantic Segmentation487
This paper presents a new framework for open-vocabulary semantic segmentation with the pre-trained vision-language model, named Side Adapter Network (SAN), which significantly outperforms other counterparts, with up to 18 times fewer trainable parameters and 19 times faster inference speed.
- CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation484
A new advanced VLA architecture derived from VLM that has a specialized action module conditioned on VLM output is presented that not only significantly surpasses existing VLAs in task performance and but also exhibits remarkable adaptation to new robots and generalization to unseen objects and backgrounds.
- Learning to Prompt for Open-Vocabulary Object Detection with Vision-Language Model482
A novel method, detection prompt (DetPro), to learn continuous prompt representations for open-vocabulary object detection based on the pre-trained vision-language model, which outperforms the baseline ViLD in all settings.
- A Simple Baseline for Open-Vocabulary Semantic Segmentation with Pre-trained Vision-Language Model360
This paper refuses the use of the prevalent one-stage FCN based framework, and advocates a two-stage semantic segmentation framework, with the first stage extracting generalizable mask proposals and the second stage leveraging an image based CLIP model to perform open-vocabulary classification on the masked image crops which are generated in the firststage.
- RAIN: Your Language Models Can Align Themselves without Finetuning189
This work introduces a novel inference method, Rewindable Auto-regressive INference (RAIN), that allows pre-trained LLMs to evaluate their own generation and use the evaluation results to guide rewind and generation for AI safety.
- Unsupervised Prompt Learning for Vision-Language Models189
This paper presents an unsupervised prompt learning (UPL) approach to avoid prompt engineering while simultaneously improving transfer performance of CLIP-like vision-language models.
- A Simple Multi-Modality Transfer Learning Baseline for Sign Language Translation169
This simple baseline surpasses the previous state-of-the-art results on two sign language translation benchmarks, demonstrating the effectiveness of transfer learning and can serve as a solid baseline for future research.
- Point-Set Anchors for Object Detection, Instance Segmentation and Pose Estimation158
This work argues that the image features extracted at a central point contain limited information for predicting distant keypoints or bounding box boundaries, due to object deformation and scale/orientation variation, and proposes to perform regression from a set of points placed at more advantageous positions to facilitate inference.
- Global Context Networks152
This paper replaces the one-layer transformation function of the non-local block by a two-layer bottleneck, which further reduces the parameter number considerably, and effectively models global context in a lightweight manner, allowing it to be applied at multiple layers of a backbone network to form a global context network (GCNet).
- AnyTool: Self-Reflective, Hierarchical Agents for Large-Scale API Calls151
AnyTool, a large language model agent designed to revolutionize the utilization of a vast array of tools in addressing user queries, is introduced, operating under the assumption that a subset of these APIs could potentially resolve the queries.
- Natural Language-Assisted Sign Language Recognition119
The Natural Language-Assisted Sign Language Recognition (NLA-SLR) framework is proposed, which exploits semantic information contained in glosses (sign labels) and introduces a novel backbone, video-keypoint network, which not only models both RGB videos and human body keypoints but also derives knowledge from sign videos of different temporal receptive fields.
- High-Fidelity and Arbitrary Face Editing89
This work proposes a simple yet effective method named HifaFace, which achieves high-fidelity and arbitrary face editing, outperforming other state-of-the-art approaches.
- Dual Path Learning for Domain Adaptation of Semantic Segmentation77
This paper proposes a novel dual path learning (DPL) framework that contains two complementary and interactive single-domain adaptation pipelines aligned in source and target domain respectively and shows superiority of the DPL model over the state-of-the-art methods.
- ADNet: Leveraging Error-Bias Towards Normal Direction in Face Alignment72
Anisotropic direction loss (ADL) and anisotropic attention module (AAM) for coordinate and heatmap regression, respectively are proposed and integrated into an optimized end-to-end training pipeline named ADNet, which achieves state-of-the-art results on 300W, WFLW and COFW datasets.
- Cross-Model Pseudo-Labeling for Semi-Supervised Action Recognition66
This work proposes a more effective pseudo-labeling scheme, called Cross-Model Pseudo-Labeling (CMPL), which introduces a lightweight auxiliary network in addition to the primary backbone, and asks them to predict pseudo-labels for each other, and observes that these two models tend to learn complementary representations from the same video clips.
- TinyMIM: An Empirical Study of Distilling MIM Pre-trained Models64
Distillation techniques to transfer the success of large MIM-based pre-trained models to smaller ones are explored, suggesting an alternative way for developing small vision Transformer models by exploring better training methods rather than introducing inductive biases into architectures as in most previous works.
- Revisiting Referring Expression Comprehension Evaluation in the Era of Large Multimodal Models62
Ref-L4 is introduced, a comprehensive REC benchmark, specifically designed to evaluate modern REC models, and several LMMs capable of handling the REC task are evaluated, showing significant accuracy improvements, thus highlighting the impact of benchmark noise.
- CiCo: Domain-Aware Sign Language Retrieval via Cross-Lingual Contrastive Learning61
This work adopts a domain-agnostic sign encoder pre-trained on large-scale sign videos into the target domain via pseudo-labeling, and develops a framework, termed as domain-aware sign language retrieval via Cross-lingual Contrastive learning or CiCo, which outperforms the pioneering method by large margins on various datasets.
- Improving Continuous Sign Language Recognition with Cross-Lingual Signs59
The feasibility of utilizing multilingual sign language corpora to facilitate monolingual CSLR is explored, and the underlying idea is to identify the cross-lingual signs in one sign language and properly leverage them as auxiliary training data to improve the recognition capability of another.
- 57
- 53
- 51
- Design and Interpretation of Universal Adversarial Patches in Face Detection51
This work investigates a phenomenon: patches designed to suppress real face detection appear face-like, which holds generally across different initialization, locations, scales of patches, backbones, and state-of-the-art face detection frameworks.
- 50
- Frame-wise Action Representations for Long Videos via Sequence Contrastive Learning48
This paper introduces a novel contrastive action representation learning (CARL) framework to learn frame-wise action representations, especially for long videos, in a self-supervised manner and shows outstanding performance on video alignment and fine-grained frame retrieval tasks.
- 43
- Conditional DETR V2: Efficient Detection Transformer with Box Queries42
Inspired by Conditional DETR, an improved DETR with fast training convergence, the object query is reformulate into the format of the box query that is a composition of the embeddings of the reference point and the transformation of thebox with respect to thereference point.
- 41
- Attentive Mask CLIP39
An attentive token removal approach, which retains a small number of tokens that have a strong semantic correlation to the corresponding text description, which outperforms original CLIP and CLIP variant with random token removal while saving the training time and enables efficient multi-view contrastive learning.
- Beyond Text: Frozen Large Language Models in Visual Signal Comprehension38
The Vision-to-Language Tokenizer is presented; abbreviated as V2T Tokenizer, which transforms an image into a “foreign language” with the combined aid of an encoder-decoder, the LLM vocabulary, and a CLIP model.
- DeepMIM: Deep Supervision for Masked Image Modeling37
This paper revisits deep supervision for masked image modeling (MIM) that pre-trains a Vision Transformer (ViT) via a mask-and-predict scheme and significantly boosts the representation capability of each layer.
- 33
- 32
- Towards Online Continuous Sign Language Recognition and Translation31
The first step towards online CSLR is taken, and the online recognition model can be extended to support online translation by integrating a gloss-to-text network and can enhance the performance of any offline model.
- Two-shot Video Object Segmentation29
This work demonstrates the feasibility of training a satisfactory VOS model on sparsely annotated videos—the authors merely require two labeled frames per training video while the performance is sustained and presents a general way to train VOS models on two-shot VOS datasets.
- AniPortraitGAN: Animatable 3D Portrait Generation from 2D Image Collections27
An animatable 3D-aware GAN that generates portrait images with controllable facial expression, head pose, and shoulder movements and a dual-camera rendering and adversarial learning scheme is proposed to improve the quality of the generated faces.
- UniGraspTransformer: Simplified Policy Distillation for Scalable Dexterous Robotic Grasping26
UniGraspTransformer is introduced, a universal Transformer-based network for dexterous robotic grasping that simplifies training while enhancing scalability and performance, and generates a broader range of grasping poses for objects in various shapes and orientations, resulting in more diverse grasp strategies.
- 26
- 25
- Boosting Zero-Shot Learning via Contrastive Optimization of Attribute Representations24
A new framework to boost ZSL by explicitly learning attribute prototypes beyond images and contrastively optimizing them with attribute-level features within images is proposed.
- 23
- 22
- Anomaly Detection via Local Coordinate Factorization and Spatio-Temporal Pyramid21
An efficient anomaly detection approach which can perform both real-time and multi-scale detection and an online method to update the local coordinates which makes the approach self-adaptive to background change which typically occurs in real-world setting is proposed.
- Rethinking Generative Large Language Model Evaluation for Semantic Comprehension18
An RWQ-Elo rating system is introduced, engaging 24 LLMs such as GPT-4, GPT-3.5, Google-Gemini-Pro and LLaMA-1/-2 in a two-player competitive format, with GPT-4 serving as the judge and each LLM receives an Elo rating thereafter.
- A Large-Scale Human-Centric Benchmark for Referring Expression Comprehension in the LMM Era15
This work presents HC-RefLoCo (Human-Centric Referring Expression Comprehension with Long Context), a benchmark that includes 13,452 images, 24,129 instances, and 44,738 detailed annotations, encompassing a vocabulary of 18,681 words, and highlights its potential to advance human-centric AI by challenging contemporary REC models with comprehensive and varied data.
- 14
- 11
- A Hong Kong Sign Language Corpus Collected from Sign-interpreted TV News11
This paper introduces TVB-HKSL-News, a new Hong Kong sign language (HKSL) dataset collected from a TV news program over a period of 7 months to support the investigation of how well large-vocabulary continuous sign language recognition/translation can be done for a single signer given a (relatively) large amount of his/her training data.
- 10
- Expressive and Generalizable Low-rank Adaptation for Large Models via Slow Cascaded Learning6
This paper presents LoRA Slow Cascade Learning (LoRASC), an innovative technique designed to enhance LoRA's expressiveness and generalization capabilities while preserving its training efficiency, and introduces a slow-fast update mechanism and cascading noisy tuning to bolster generalization.
- 5
- Modeling Multi-modal Cross-interaction for Multi-label Few-shot Image Classification Based on Local Feature Selection3
A strategy in which label prototypes are gradually refined using word embeddings and a Loss Change Measurement strategy to select the local features from the training images that are most likely to be representative of a given label.
- 3
- Exploring Non-additive Randomness on ViT against Query-Based Black-Box Attacks2
This work first taxonomize the stochastic defense strategies against QBBA, and proposes to explore non-additive randomness in models to defend against it, focusing on underexplored Vision Transformers based on their flexible architectures.
- Ink3D: Sculpting 3D Assets with Extremely Complex Textures via Video Generative Models1
This work introduces Ink3D, a framework that bridges 3D generation with large-scale video generative models to synthesize extremely complex textures and introduces TextureOptimizer, a neural baking module that integrates dense multi-view observations while mitigating geometry inconsistencies arising from video generation.
- 1
- LoRASC: Expressive and Generalizable Low-rank Adaptation for Large Models via Slow Cascaded Learning1
- 1
- Anomaly Detection with ELM-Based Visual Attribute and Spatio-temporal Pyramid1
A novel approach for automatic anomaly detection in surveillance video is proposed which can cope with multi-scale and complicated events, and uses the Spatio-temporal Pyramid to capture the spatial and temporal continuity of an anomalies event.
- –
- –
- –
- –
- InsightTok: Improving Text and Face Fidelity in Discrete Tokenization for Autoregressive Image Generation–
InsightTok is proposed, a simple yet effective discrete visual tokenization framework that enhances text and face fidelity through localized, content-aware perceptual losses and significantly outperforms prior tokenizers in text and face reconstruction without compromising general reconstruction quality.
- –
- –
- –
- Manifold-Regularized Selectable Factor Extraction for Semi-supervised Image Classification–
This work proposes a novel manifold regularized SFE (MRSFE) framework for semi-supervised image classification that could select and extract reliable discriminative features when the labeled samples are scarce and an efficient and easy-to-implement algorithm is designed to find the solutions.
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.