Chang Xu

116

Papers

400

Total Citations

Papers (116)

CNNpack: Packing Convolutional Neural Networks in the Frequency Domain

Lumina-Image 2.0: A Unified and Efficient Image Generative Framework

VSSD: Vision Mamba with Non-Causal State Space Duality

Hybrid Proposal Refiner: Revisiting DETR Series from the Faster R-CNN Perspective

TimeDP: Learning to Generate Multi-Domain Time Series with Domain Prompts

Data-efficient Large Vision Models through Sequential Autoregression

Harnessing Vision Foundation Models for High-Performance, Training-Free Open Vocabulary Segmentation

Training-free Composite Scene Generation for Layout-to-Image Synthesis

Learning Visual Abstract Reasoning through Dual-Stream Networks

SMARTIES: Spectrum-Aware Multi-Sensor Auto-Encoder for Remote Sensing Images

Adapt without Forgetting: Distill Proximity from Dual Teachers in Vision-Language Models

Enhancing Privacy-Utility Trade-offs to Mitigate Memorization in Diffusion Models

WaterDiffusion: Learning a Prior-involved Unrolling Diffusion for Joint Underwater Saliency Detection and Visual Restoration

Anti-Exposure Bias in Diffusion Models

Feature Clipping for Uncertainty Calibration

Learning Mask Invariant Mutual Information for Masked Image Modeling

Harnessing Edge Information for Improved Robustness in Vision Transformers

Position: AI Scaling: From Up to Down and Out

Uncertainty Weighted Gradients for Model Calibration

Residual Learning in Diffusion Models

Beyond One-Hot Labels: Semantic Mixing for Model Calibration

Efficient Image-to-Image Diffusion Classifier for Adversarial Robustness

ReNAS: Relativistic Evaluation of Neural Architecture Search

Learning Student Networks in the Wild

AdderSR: Towards Energy Efficient Image Super-Resolution

Pre-Trained Image Processing Transformer

Distilling Object Detectors via Decoupled Features

Positive-Unlabeled Data Purification in the Wild for Object Detection

HourNAS: Extremely Fast Neural Architecture Search Through an Hourglass Lens

Prioritized Architecture Sampling With Monto-Carlo Tree Search

Where and What? Examining Interpretable Disentangled Representations

Manifold Regularized Dynamic Network Pruning

BCNet: Searching for Network Width With Bilaterally Coupled Network

DeepFake Disrupter: The Detector of DeepFake Is My Friend

SimMatch: Semi-Supervised Learning With Similarity Matching

Hire-MLP: Vision MLP via Hierarchical Rearrangement

DyRep: Bootstrapping Training With Dynamic Re-Parameterization

GreedyNASv2: Greedier Search With a Greedy Path Filter

An Image Patch Is a Wave: Phase-Aware Vision MLP

Patch Slimming for Efficient Vision Transformers

CMT: Convolutional Neural Networks Meet Vision Transformers

Dynamic Coarse-To-Fine Learning for Oriented Tiny Object Detection

PSVT: End-to-End Multi-Person 3D Pose and Shape Estimation With Progressive Video Transformers

Trade-Off Between Robustness and Accuracy of Vision Transformers

Private Image Generation With Dual-Purpose Auxiliary Classifier

Adversarial Robustness via Random Projection Filters

Co-Evolutionary Compression for Unpaired Image Translation

Self-Supervised Representation Learning From Multi-Domain Data

Approximated Bilinear Modules for Temporal Modeling

Data-Free Learning of Student Networks

Learning Frequency-Aware Dynamic Network for Efficient Super-Resolution

Weakly Supervised Contrastive Learning

Personalized Image Generation for Color Vision Deficiency Population

What Can Simple Arithmetic Operations Do for Temporal Modeling?

Re-mine, Learn and Reason: Exploring the Cross-modal Semantic Correlations for Language-guided HOI detection

Speech4Mesh: Speech-Assisted Monocular 3D Facial Reconstruction for Speech-Driven 3D Facial Animation

SimMatchV2: Semi-Supervised Learning with Graph Consistency

Diffusion Action Segmentation

Optical Flow Distillation: Towards Efficient and Stable Video Style Transfer

Learning Disentangled Representations with Latent Variation Predictability

On Dropping Clusters to Regularize Graph Convolutional Neural Networks

One Size Does NOT Fit All: Data-Adaptive Adversarial Training

RFLA: Gaussian Receptive Field Based Label Assignment for Tiny Object Detection

ViTAS: Vision Transformer Architecture Search

GeminiFusion: Efficient Pixel-wise Multimodal Fusion for Vision Transformer

Minimizing Labeled, Maximizing Unlabeled: An Image-Driven Approach for Video Instance Segmentation

UniGraspTransformer: Simplified Policy Distillation for Scalable Dexterous Robotic Grasping

EfficientVMamba: Atrous Selective Scan for Light Weight Visual Mamba

AMD: Autoregressive Motion Diffusion

Random Entangled Tokens for Adversarially Robust Vision Transformer

Towards Memorization-Free Diffusion Models

Adversarial Robustness via Deformable Convolution with Stochasticity

Imitation Learning from Purified Demonstrations

Bridging Data Gaps in Diffusion Models with Adversarial Noise-Based Transfer Learning

Detecting Any instruction-to-answer interaction relationship:Universal Instruction-to-Answer Navigator for Med-VQA

Self-Supervised Representation Learning by Rotation Feature Decoupling

Image-Question-Answer Synergistic Network for Visual Dialog

GhostNet: More Features From Cheap Operations

Frequency Domain Compact 3D Convolutional Neural Networks

Neuromorphic Camera Guided High Dynamic Range Imaging

A Semi-Supervised Assessor of Neural Architectures

CARS: Continuous Evolution for Efficient Neural Architecture Search

AdderNet: Do We Really Need Multiplications in Deep Learning?

On Positive-Unlabeled Classification in GAN

Context Aware Graph Convolution for Skeleton-Based Action Recognition

Hit-Detector: Hierarchical Trinity Architecture Search for Object Detection

Learning Versatile Filters for Efficient Convolutional Neural Networks

Learning from Bad Data via Generation

Positive-Unlabeled Compression on the Cloud

Adapting Neural Architectures Between Domains

UnModNet: Learning to Unwrap a Modulo Image for High Dynamic Range Imaging

Searching for Low-Bit Weights in Quantized Neural Networks

SCOP: Scientific Control for Reliable Neural Network Pruning

Kernel Based Progressive Distillation for Adder Neural Networks

ReSSL: Relational Self-Supervised Learning with Weak Augmentation

An Empirical Study of Adder Neural Networks for Object Detection

Towards Stable and Robust AdderNets

Augmented Shortcuts for Vision Transformers

Handling Long-tailed Feature Distribution in AdderNets

Learning Frequency Domain Approximation for Binary Neural Networks

Neural Architecture Dilation for Adversarial Robustness

GhostNetV2: Enhance Cheap Operation with Long-Range Attention

Searching for Better Spatio-temporal Alignment in Few-Shot Action Recognition

Random Normalization Aggregation for Adversarial Defense

Knowledge Distillation from A Stronger Teacher

Detecting Any Human-Object Interaction Relationship: Universal HOI Detector with Spatial Prompt Learning on Foundation Models

Revisit the Power of Vanilla Knowledge Distillation: from Small Scale to Large Scale

Adversarial Robustness through Random Weight Sampling

Rethinking Conditional Diffusion Sampling with Progressive Guidance

Beyond Pretrained Features: Noisy Image Modeling Provides Adversarial Defense

Stable Diffusion is Unstable

Knowledge Diffusion for Distillation

Contrastive Sampling Chains in Diffusion Models

One-for-All: Bridge the Gap Between Heterogeneous Architectures in Knowledge Distillation

Beyond Filters: Compact Feature Map for Portable Deep Model

LegoNet: Efficient Convolutional Neural Networks with Lego Filters