Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks345 from public data
- Effective Whole-body Pose Estimation with Two-stages Distillation446
This work presents a two-stage pose Distillation for Whole-body Pose estimators, named DWPose, to improve their effectiveness and efficiency and releases a series of models with different sizes, from tiny to large, for satisfying various downstream tasks.
- Focal and Global Knowledge Distillation for Detectors432
Focal and Global Distillation (FGD) is proposed, which separates the foreground and background, forcing the student to focus on the teacher's critical pixels and channels, and compensating for missing global information in focal distillation.
- Self-Attention ConvLSTM for Spatiotemporal Prediction384
A novel self-attention memory (SAM) is proposed to memorize features with long-range dependencies in terms of spatial and temporal domains and is applied to perform frame prediction on the MovingMNIST and KTH datasets and traffic flow Prediction on the TexiBJ dataset.
- 265
- Differentially Private Data Release through Multidimensional Partitioning193
This paper proposes two multidimensional partitioning strategies including a baseline cell-based partitioning and an innovative kd-tree based partitioning for differentially private histogram release based on an interactive differential privacy interface.
- Automated known problem diagnosis with event traces188
This work proposes to use system behavior information such as system event traces to build correlations with solved problems, instead of using only vague text descriptions as in existing practices to enable automatic identification of the root cause of a problem if it is a known one, which would further lead to its resolution.
- Strider: a black-box, state-based approach to change and configuration management and support184
Troubleshooting of configuration failures are used to demonstrate that the Strider approach reduces problem complexity by several orders of magnitude, making root-cause analysis possible.
- From Knowledge Distillation to Self-Knowledge Distillation: A Unified Approach with Normalized Loss and Customized Soft Labels166
This work unifies the formulations of the two tasks by decomposing and reorganizing the generic KD loss into a Normalized KD (NKD) loss and customized soft labels for both target class (image’s category) and non-target classes named Universal Self-KD (USKD).
- ColorNet: Investigating the Importance of Color Spaces for Image Classification160
This model essentially takes an RGB image as input, simultaneously converts the image into 7 different color spaces and uses these as inputs to individual densenets to reduce computation overhead and number of hyperparameters required.
- Low-Rank Preserving Projections158
This paper proposes a novel dimensionality reduction method, named low-rank preserving projections (LRPP) for image classification, and shows the effectiveness and the feasibility of the proposed method with encouraging results.
- 147
- FlatQuant: Flatness Matters for LLM Quantization144
This paper proposes FlatQuant (Fast and Learnable Affine Transformation), a new post-training quantization approach that enhances the flatness of weights and activations, and identifies optimal affine transformations for each linear layer, calibrated in hours via a lightweight objective.
- ChartBench: A Benchmark for Complex Visual Reasoning in Charts127
ChartBench, a comprehensive benchmark designed to assess chart comprehension and data reliability through complex visual reasoning, is proposed and two baselines based on the chain of thought and supervised fine-tuning are proposed to improve model performance on unannotated charts.
- Convolution Meets LoRA: Parameter Efficient Finetuning for Segment Anything Model121
Conv-LoRA not only preserves SAM's extensive segmentation knowledge but also revives its capacity of learning high-level image semantics, which is constrained by SAM's foreground-background segmentation pretraining.
- CricaVPR: Cross-Image Correlation-Aware Representation Learning for Visual Place Recognition115
A robust global representation method with cross-image correlation awareness for VPR, named CricaVPR, which uses the attention mechanism to correlate multiple images within a batch and introduces the multi-scale local information to further enhance the cross-image correlation-aware representation.
- DreamDiffusion: Generating High-Quality Images from Brain EEG Signals111
Quantitative and qualitative results demonstrate the effectiveness of the proposed DreamDiffusion method as a significant step towards portable and low-cost ``thoughts-to-image'', with potential applications in neuroscience and computer vision.
- Towards Seamless Adaptation of Pre-trained Models for Visual Place Recognition100
A hybrid adaptation method to achieve both global and local adaptation efficiently, in which only lightweight adapters are tuned without adjusting the pre-trained model, and a mutual nearest neighbor local feature loss, which ensures proper dense local features are produced for local matching and avoids time-consuming spatial verification in re-ranking.
- Learning to Run Challenge Solutions: Adapting Reinforcement Learning Methods for Neuromusculoskeletal Environments98
This work presents eight solutions that used deep reinforcement learning approaches, based on algorithms such as Deep Deterministic Policy Gradient, Proximal Policy Optimization, and Trust Region Policyoptimization, to make it run as fast as possible through an obstacle course.
- 90
- SparseFlex: High-Resolution and Arbitrary-Topology 3D Shape Modeling89
This paper introduces SparseFlex, a novel sparse-structured isosurface representation that enables differentiable mesh reconstruction at resolutions up to 10243 directly from rendering losses, and introduces a frustum-aware sectional voxel training strategy that activates only relevant voxels during rendering, dramatically reducing memory consumption and enabling high-resolution training.
- Structurally Incoherent Low-Rank Nonnegative Matrix Factorization for Image Classification86
This paper proposes a novel nonnegative factorization method, called structurally incoherent low-rank NMF (SILR-NMF), in which they jointly consider structural incoherence and low- rank properties of data for image classification.
- Quantization Hurts Reasoning? An Empirical Study on Quantized Reasoning Models83
This paper conducts the first systematic study on quantized reasoning models, evaluating the open-sourced DeepSeek-R1-Distilled Qwen and LLaMA families ranging from 1.5B to 70B parameters, QwQ-32B, and Qwen3-8B using state-of-the-art algorithms at varying bit-widths.
- Learning Imbalanced Data with Vision Transformers80
This paper systematically investigates the ViTs' performance in LTR and proposes LiVT to train ViTs from scratch only with LT data, and derives the unbiased extension of Sigmoid and compensates extra logit margins for deploying it to ameliorate its performance.
- Uni-Perceiver v2: A Generalist Model for Large-Scale Vision and Vision-Language Tasks75
This paper proposes Uni-Perceiver v2, which is the first generalist model capable of handling major large-scale vision and vision-language tasks with competitive performance and proposes an effective optimization technique named Task-Balanced Gradient Normalization to ensure stable multi-task learning with an unmixed sampling strategy, helpful for tasks requiring large batch-size training.
- UPop: Unified and Progressive Pruning for Compressing Vision-Language Transformers71
The UPop framework is proposed as a universal vison-language Transformer compression framework, which incorporates unifiedly searching multimodal subnets in a continuous optimization space from the original model, which enables automatic assignment of pruning ratios among compressible modalities and structures.
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 2026-10-12. 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.