Haoyu Wang (王昊宇)

JoDiffusion: Jointly Diffusing Image with Pixel-Level Annotations for Semantic Segmentation Promotion
Haoyu Wang, Lei Zhang, Wenrui Liu, Dengyang Jiang, Wei Wei, Chen Ding

Code | Models

• A joint diffusion framework for semantic segmentation dataset generation using an annotation VAE for latent alignment and a boundary mode-based mask optimization strategy.

No Other Representation Component Is Needed: Diffusion Transformers Can Provide Representation Guidance by Themselves
Dengyang Jiang, Mengmeng Wang, Liuzhuozheng Li, Lei Zhang, Haoyu Wang, Wei Wei, Guang Dai, Yanning Zhang, Jingdong Wang

Project | Code

• Enhances Diffusion Transformers’ representation and generation through self-representation alignment via self-distillation, eliminating external components.

Prompt-Free Conditional Diffusion for Multi-object Image Augmentation
Haoyu Wang, Lei Zhang, Wei Wei, Chen Ding, Yanning Zhang

Code | Models

• A framework for multi-object image augmentation using local-global semantic fusion and a reward model-based counting loss.

Low-Biased General Annotated Dataset Generation
Dengyang Jiang*, Haoyu Wang*, Lei Zhang, Wei Wei, Guang Dai, Mengmeng Wang, Jingdong Wang, Yanning Zhang

Code

• A framework generating low-biased annotated datasets using a fine-tuned diffusion model with bi-level semantic alignment and quality assurance for enhanced backbone generalization.

Adapt Anything: Tailor Any Image Classifier across Domains And Categories Using Text-to-Image Diffusion Models
Weijie Chen*, Haoyu Wang*, Shicai Yang, Lei Zhang, Wei Wei, Yanning Zhang, Luojun Lin, Di Xie, Yueting Zhuang

Paper

• Uses text-to-image diffusion models to create synthetic data, enabling image classifier adaptation across domains and categories without real-world source data.

Glocal energy-based learning for few-shot open-set recognition
Haoyu Wang*, Guansong Pang*, Peng Wang*, Lei Zhang, Wei Wei, Yanning Zhang

Code

• A novel energy-based model for few-shot open-set recognition using global and local features.