Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks116 from public data
- A Survey on Vision Transformer4,578
This paper reviews these vision transformer models by categorizing them in different tasks and analyzing their advantages and disadvantages, and takes a brief look at the self-attention mechanism in computer vision, as it is the base component in transformer.
- CMT: Convolutional Neural Networks Meet Vision Transformers992
A new transformer based hybrid network is proposed by taking advantage of transformers to capture long-range dependencies, and of CNNs to extract local information, obtaining much better trade-off for accuracy and efficiency than previous CNN-based and transformer-based models.
- Multimodal Token Fusion for Vision Transformers327
The design of TokenFusion allows the transformer to learn correlations among multimodal features, while the single-modal transformer architecture remains largely intact, and surpasses state-of-the-art methods in three typical vision tasks: multimodals image-to-image translation, RGB-depth semantic segmentation, and 3D object detection with point cloud and images.
- Distilling Object Detectors via Decoupled Features278
This paper presents a novel distillation algorithm via decoupled features (DeFeat) for learning a better student detector that is able to surpass the state-of-the-art distillation methods for object detection.
- CARS: Continuous Evolution for Efficient Neural Architecture Search275
This work develops an efficient continuous evolutionary approach for searching neural networks that provides a series of networks with the number of parameters ranging from 3.7M to 5.1M under mobile settings and surpasses those produced by the state-of-the-art methods on the benchmark ImageNet dataset.
- 217
- Pose guided structured region ensemble network for cascaded hand pose estimation208
A novel approach named as pose guided structured region ensemble network (Pose-REN), which extracts regions from the feature maps of convolutional neural network and generates more optimal and representative features for hand pose estimation.
- 3D Hand Pose Estimation: From Current Achievements to Future Goals194
This paper investigates 11 state-of-the-art methods on three tasks: single frame 3D pose estimation, 3D hand tracking, and hand pose estimation during object interaction.
- Spatial-Temporal Attention Res-TCN for Skeleton-Based Dynamic Hand Gesture Recognition149
An end-to-end Spatial-Temporal Attention Residual Temporal Convolutional Network (STA-Res-TCN) for skeleton-based dynamic hand gesture recognition, which learns different levels of attention and assigns them to each spatial-temporal feature extracted by the convolution filters at each time step.
- 140
- Hire-MLP: Vision MLP via Hierarchical Rearrangement128
Hire-MLP is presented, a simple yet competitive vision MLP architecture via Hierarchical rearrangement, which contains two levels of rearrangements to enable information communication between different regions and capture global context by circularly shifting all tokens along spatial directions.
- Hit-Detector: Hierarchical Trinity Architecture Search for Object Detection110
This work proposes a hierarchical trinity search framework to simultaneously discover efficient architectures for all components of object detector in an end-to-end manner and empirically reveals that different parts of the detector prefer different operators.
- SHPR-Net: Deep Semantic Hand Pose Regression From Point Clouds110
A deep semantic hand pose regression network (SHPR-Net) for hand pose estimation from point sets, which consists of two subnetworks: a semantic segmentation subnetwork and a hand poses regression subnetwork, which is more robust to geometric transformations.
- Region ensemble network: Towards good practices for deep 3D hand pose estimation107
A Region Ensemble Network for directly 3D coordinate regression that significantly improves the performance of ConvNet for hand pose estimation and achieves strong performance on par or better than state-of-the-art algorithms on three public hand pose datasets.
- 97
- 90
- MFA-Net: Motion Feature Augmented Network for Dynamic Hand Gesture Recognition from Skeletal Data85
A novel motion feature augmented network (MFA-Net) is proposed for dynamic hand gesture recognition from skeletal data that exploits motion features of finger and global movements to augment features of deep network for gesture recognition.
- GeminiFusion: Efficient Pixel-wise Multimodal Fusion for Vision Transformer82
This paper first critiques prior token exchange methods which replace less informative tokens with inter-modal features, and demonstrates exchange based methods underperform cross-attention mechanisms, while the computational demand of the latter inevitably restricts its use with longer sequences.
- 69
- Pangu Embedded: An Efficient Dual-system LLM Reasoner with Metacognition68
Experimental results on benchmarks including AIME 2024, GPQA, and LiveCodeBench demonstrate that Pangu Embedded with 7B parameters, outperforms similar-size models like Qwen3-8B and GLM4-9B.
- PPT: Token Pruning and Pooling for Efficient Vision Transformers52
This paper proposes a novel acceleration framework, namely token Pruning&Pooling Transformers (PPT), to adaptively tackle these two types of redundancy in different layers of ViTs by heuristically integrating both token pruning and token pooling techniques in ViTs without additional trainable parameters.
- SLAB: Efficient Transformers with Simplified Linear Attention and Progressive Re-parameterized Batch Normalization51
A novel method named PRepBN is proposed to progressively replace LayerNorm with re-parameterized BatchNorm in training and a simplified linear attention (SLA) module is proposed that is simple yet effective to achieve strong performance.
- 44
- 41
- 40
- TimePro: Efficient Multivariate Long-term Time Series Forecasting with Variable- and Time-Aware Hyper-state38
TimePro is an innovative Mamba-based model that constructs variate- and time-aware hyper-states that can perceive both variable relationships and salient temporal information, which helps the model make accurate forecasting.
- 34
- SlimLLM: Accurate Structured Pruning for Large Language Models30
This paper proposes an effective and fast structured pruning method named SlimLLM for large language models that outperforms other methods and achieves state-of-the-art performance on the LLaMA benchmark results.
- Multimodal Language Models See Better When They Look Shallower28
This work offers the first principled study of visual layer selection in MLLMs, showing that MLLMs can often see better when they look shallower, and proposes a lightweight feature fusion method that strategically incorporates shallower layers, achieving consistent improvements over both single-layer and specialized fusion baselines.
- 28
- GenVidBench: A 6-Million Benchmark for AI-Generated Video Detection25
GenVidBench is introduced, a challenging AI-generated video detection dataset with several key advantages that can efficiently develop and evaluate AI-generated video detection models and presents extensive experimental results with advanced video classification models.
- AirV2X: Unified Air-Ground Vehicle-to-Everything Collaboration23
The AirV2X-Perception dataset facilitates the development and standardized evaluation of Vehicle-to-Drone (V2D) algorithms, addressing a critical gap in the rapidly expanding field of aerial-assisted autonomous driving systems.
- Efficient Branch and Bound Search with Application to Computer-Aided Design22
The author explains the development of Justification Equivalence in the Finite State Space and its role in Logic Verification via Test Generation.
- Pangu Pro MoE: Mixture of Grouped Experts for Efficient Sparsity20
Mixture of Grouped Experts (MoGE), which groups the experts during selection and balances the expert workload better than MoE in nature, is introduced, which leads to better expert load balancing and more efficient execution for both model training and inference on Ascend NPUs.
- MTP: Multi-Task Pruning for Efficient Semantic Segmentation Networks20
To identify the redundancy in segmentation networks, a multi-task channel pruning approach is presented and the superiority of the proposed al-gorithm over the state-of-the-art pruning methods is illustrated.
- Notice of Removal Two-stream convolutional neural network for accurate RGB-D fingertip detection using depth and edge information20
This paper presents a novel two-stream convolutional neural network (CNN) for RGB-D fingertip detection, which outperforms state-of-the-art fingertip Detection methods on the public dataset HandNet with an average 3D error of 9.9mm and shows comparable accuracy of fingertip estimation on NYU hand dataset.
- HourNAS: Extremely Fast Neural Architecture Search Through an Hourglass Lens19
An hourglass-inspired approach (HourNAS) for extremely fast NAS to identify the vital blocks and make them the priority in the architecture search, which outperforms the state-of-the-art methods.
- 16
- ExCP: Extreme LLM Checkpoint Compression via Weight-Momentum Joint Shrinking15
A novel Extreme Checkpoint Compression framework, which significantly reduces the required storage of training checkpoints while achieving nearly lossless performance, and proposes a weight-momentum joint shrinking method to utilize another important information during the model optimization.
- Multi-Task Pruning for Semantic Segmentation Networks15
This paper presents a multi-task channel pruning approach for semantic segmentation networks, and develops an alternative scheme for optimizing importance scores of filters in the entire network.
- 13
- Teaching Solar Energy Systems Design using Game-Based Virtual Reality12
The approach enables students to immerse themselves in a virtual environment, which is safe for both students and their teachers, and improved student awareness of solar energy systems and how they can be used to satisfy domestic energy demands.
- Bi-Stream Pose-Guided Region Ensemble Network for Fingertip Localization From Stereo Images12
A novel scheme termed bi-stream pose-guided region ensemble network (Bi-Pose-REN), which extracts more representative feature regions around joints in the feature maps under the guidance of the previously estimated pose, to better present the performance of fingertip localization on THU-Bi-Hand.
- Static hand gesture recognition based on finger root-center-angle and length weighted Mahalanobis distance11
A new framework for static hand gesture recognition is presented, including a novel and discriminative features called root-center-angles to alleviate the influence of the variations of gestures and a distance metric called finger length weighted Mahalanobis distance (FLWMD) to measure the dissimilarity of the hand gestures.
- 10
- Single Domain Generalization for Few-Shot Counting via Universal Representation Matching9
This work finds that incorporating universal vision-language representations distilled from a large scale pretrained vision-language model into the correlation construction process substantially improves robustness to domain shifts without compromising in domain performance.
- Fundamentals of CMOS design9
This chapter discusses a few basic and important concepts of complementary metal oxide semiconductor (CMOS) technology to aid in the learning process and facilitates greater understanding of the electronic design automation (EDA) subjects.
- 8
- Sequential Circuit Test Generation8
In general, digital circuits with storage capability, or “memory”, are called sequential circuits, while digital circuits without storage capability are called combinational circuits.
- 7
- 7
- Beyond Textual CoT: Interleaved Text-Image Chains with Deep Confidence Reasoning for Image Editing6
The proposed Multimodal Reasoning Edit (MURE), a novel framework that shifts the visual editing process from purely text-based reasoning to a series of interleaved textual and visual rationales, and introduces Multimodal Deep Confidence (MMDC) reasoning paradigm.
- VLM-Pruner: Buffering for Spatial Sparsity in an Efficient VLM Centrifugal Token Pruning Paradigm6
This work introduces a centrifugal token pruning paradigm that enables near-to-far selection while prioritizing the preservation of fine-grained object details, and designs a Buffering for Spatial Sparsity criterion that defers the selection of spatially distant tokens.
- 6
- A heuristics pulse algorithm with relaxation pruning strategy for resources re-initialized UAV path planing6
This paper focuses on solving the UAV path planning problem under the constraint of system positioning error, and a relaxed pruning strategy is introduced into the traditional Pulse algorithm to reduce the search space and avoid more redundant calculations on unfavorable scalable nodes by the proposed heuristic search strategy.
- 6
- 6
- Full-Stage Pseudo Label Quality Enhancement for Weakly-Supervised Temporal Action Localization5
A set of simple yet efficient mechanisms for pseudo label quality enhancement to build the FuSTAL framework, which decomposes the learning process into three stages and enhances pseudo label quality at each one and achieves an average mAP of 50.8% on the benchmark data THUMOS’14, outperforming the previous best method by 1.2%.
- 5
- 5
- 5
- A Binary-Classification Method Based on Dictionary Learning and ADMM for Network Intrusion Detection5
- 5
- 4
- 4
- Pangu Light: Weight Re-Initialization for Pruning and Accelerating LLMs3
Pangu Light is introduced, a framework for LLM acceleration centered around structured pruning coupled with novel weight re-initialization techniques designed to address this ``missing piece''.
- 3
- 3
- Interactive Hand Pose Estimation: Boosting accuracy in localizing extended finger joints3
A novel method to refine stretching-out finger joint locations after obtaining rough hand pose estimation is proposed, which first detects which fingers are stretching out, then neighbor pixels of certain joint vote for its new location based on random forests.
- Two-stream binocular network: Accurate near field finger detection based on binocular images3
A new framework is proposed, named Two-Stream Binocular Network (TSBnet) to detect fingertips from binocular images directly, which first shares convolutional layers for low level features of right and left images, and extracts high level features in two-stream Convolutional networks separately.
- 3
- Sequential circuit test generation using dynamic justification equivalence3
This work proposes a new technique to improve sequential circuit ATPG efficiency by focusing on not repeating previous searches, and describes the details of justification equivalence and its application in ATPG accompanied with step-by-step examples.
- Towards Lossless Ultimate Vision Token Compression for VLMs2
Lossless Ultimate Vision tokens Compression (LUVC) systematically compresses visual tokens until complete elimination at the final layer of LLM, so that the high-dimensional visual features are gradually fused into the multimodal queries.
- 2
- Blurring-Effect-Free CNN Network of Structural Edge for Focus Stacking2
This work proposes a novel convolutional neural network (BEF-CNN) to restore blurring-effect-free image patches in order to enhance all-in-focus performance and is the first work to utilize CNN to generate all- in-focus image directly instead of pixel-to-pixel correspondence with depthmap.
- Stable and real-time hand gesture recognition based on RGB-D data2
A coarse-to-fine model is proposed, which utilizes the skin color and helps to extract the hand region robustly and accurately and the principal component analysis (PCA) algorithm is introduced to estimate and then compensate the direction.
- 2
- 2
- 2
- 1
- Ada-MoGE: Adaptive Mixture of Gaussian Expert Model for Time Series Forecasting1
Ada-MoGE, an adaptive Gaussian Mixture of Experts model that integrates spectral intensity and frequency response to adaptively determine the number of experts, ensuring alignment with the input data's frequency distribution, is proposed.
- 1
- 1
- 1
- 1
- A Directional Heuristics Pulse Algorithm for a Two Resources Constrained Shortest Path Problem with Reinitialization1
A mixed-integer linear programming (MILP) model is proposed to solve the resource constrained shortest path problem (RCSPP-R) with reinitialization, and its resource accumulation can be reinitialized to zero at the reset node.
- 1
- 1
- Bi-stream Region Ensemble Network: Promoting Accuracy in Fingertip Localization from Stereo Images.1
This paper proposes a novel scheme named Bi-stream Region Ensemble Network (Bi-REN) to estimate the locations of fingertips and the wrist from stereo images directly, without converting them into depth images.
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- A Method of Speeding Image Rendering based on OpenGL–
A new method which set GPU and CPU together to working for image rendering through reorganizing the OpenGL rendering procedure is proposed and the proof of the experimentation show that this method can increasing the speed of the image rendering.
- –
- Based on Wireless Local Area Network Distributed Rendering Cluster–
This article presents the distributed render base on LAN, a Solution base on WLAN distributed render that is practical, realization and given to the student.
- –
- –
- Justification in Finite State Space–
This chapter presents a method that dynamically gathers justification information while traversing the finite state space and is used in subsequent justifications to avoid previously-explored state justification decisions.
- –
- –
- –
- The Sest Algorithm–
This chapter describes the SEST automatic test pattern generation (ATPG) algorithm, which applies the justification equivalence theorems discussed in Chapters 2 and 3.
- –
- –
- –
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.