Chen Wei

Affiliations

Johns Hopkins University

papers

1,771

total citations

papers (18)

Masked Feature Prediction for Self-Supervised Visual Pre-Training

CVPR 2022arXiv

801

citations

CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning

CVPR 2021arXiv

313

citations

Perception Encoder: The best visual embeddings are not at the output of the network

NEURIPS 2025arXiv

129

citations

SynBody: Synthetic Dataset with Layered Human Models for 3D Human Perception and Modeling

ICCV 2023arXiv

citations

Masked Autoencoders Enable Efficient Knowledge Distillers

CVPR 2023arXiv

citations

Digital Life Project: Autonomous 3D Characters with Social Intelligence

CVPR 2024arXiv

citations

AiOS: All-in-One-Stage Expressive Human Pose and Shape Estimation

CVPR 2024arXiv

citations

CP2: Copy-Paste Contrastive Pretraining for Semantic Segmentation

ECCV 2022arXiv

citations

SMAUG: Sparse Masked Autoencoder for Efficient Video-Language Pre-Training

ICCV 2023arXiv

citations

De-Diffusion Makes Text a Strong Cross-Modal Interface

CVPR 2024arXiv

citations

Growing a Twig to Accelerate Large Vision-Language Models

ICCV 2025arXiv

citations

DH-FaceVid-1K: A Large-Scale High-Quality Dataset for Face Video Generation

ICCV 2025arXiv

citations

Vision‑Language‑Vision Auto‑Encoder: Scalable Knowledge Distillation from Diffusion Models

NEURIPS 2025arXiv

citations

In Defense of Image Pre-training for Spatiotemporal Recognition

ECCV 2022arXiv

citations

QR-LoRA: Efficient and Disentangled Fine-tuning via QR Decomposition for Customized Generation

ICCV 2025arXiv

citations

Chen Wei

Affiliations

papers (18)

Masked Feature Prediction for Self-Supervised Visual Pre-Training

CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning

Perception Encoder: The best visual embeddings are not at the output of the network

Diffusion Models as Masked Autoencoders

VideoDPO: Omni-Preference Alignment for Video Diffusion Generation

Towards Generalizable Tumor Synthesis

SynBody: Synthetic Dataset with Layered Human Models for 3D Human Perception and Modeling

Masked Autoencoders Enable Efficient Knowledge Distillers

Digital Life Project: Autonomous 3D Characters with Social Intelligence

AiOS: All-in-One-Stage Expressive Human Pose and Shape Estimation

CP2: Copy-Paste Contrastive Pretraining for Semantic Segmentation

SMAUG: Sparse Masked Autoencoder for Efficient Video-Language Pre-Training

De-Diffusion Makes Text a Strong Cross-Modal Interface

Growing a Twig to Accelerate Large Vision-Language Models

DH-FaceVid-1K: A Large-Scale High-Quality Dataset for Face Video Generation

Vision‑Language‑Vision Auto‑Encoder: Scalable Knowledge Distillation from Diffusion Models

In Defense of Image Pre-training for Spatiotemporal Recognition

QR-LoRA: Efficient and Disentangled Fine-tuning via QR Decomposition for Customized Generation

papers (18)

Masked Feature Prediction for Self-Supervised Visual Pre-Training

CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning

Perception Encoder: The best visual embeddings are not at the output of the network

Diffusion Models as Masked Autoencoders

VideoDPO: Omni-Preference Alignment for Video Diffusion Generation

Towards Generalizable Tumor Synthesis

SynBody: Synthetic Dataset with Layered Human Models for 3D Human Perception and Modeling

Masked Autoencoders Enable Efficient Knowledge Distillers

Digital Life Project: Autonomous 3D Characters with Social Intelligence

AiOS: All-in-One-Stage Expressive Human Pose and Shape Estimation

CP2: Copy-Paste Contrastive Pretraining for Semantic Segmentation

SMAUG: Sparse Masked Autoencoder for Efficient Video-Language Pre-Training

De-Diffusion Makes Text a Strong Cross-Modal Interface

Growing a Twig to Accelerate Large Vision-Language Models

DH-FaceVid-1K: A Large-Scale High-Quality Dataset for Face Video Generation

Vision‑Language‑Vision Auto‑Encoder: Scalable Knowledge Distillation from Diffusion Models

In Defense of Image Pre-training for Spatiotemporal Recognition

QR-LoRA: Efficient and Disentangled Fine-tuning via QR Decomposition for Customized Generation