Leave the world a little different.
I'm a Ph.D. candidate at HCP Lab, Sun Yat-sen University, and Pengcheng Laboratory, advised by Prof. Xiaodan Liang and Assoc. Prof. Xiangyuan Lan. Before that, I received my Master's degree from UCAS and CASIA, advised by Prof. Jing Liu, and my Bachelor's degree from BJTU.
My research centers on open-ended computer vision and multimodal large language models, with first-author work on unified image/video segmentation, open-vocabulary detection, and temporal pixel-level understanding. I'm increasingly exploring multimodal agentic models.
I'm actively looking for research positions in industry and am open to collaborations on interesting
projects. If you have suitable opportunities or would like to work together, reach me at
wanghao9610@gmail.com or on WeChat
wangh9610.
A unified segmentation MLLM that extends any-segmentation from images to videos, supporting conversational instructions and visual prompts through Mask Memory for temporally consistent pixel-level perception.
A unified multimodal large language model framework that extends segment-anything capabilities to any segmentation, enabling instruction-driven interaction and pixel-level perceptual understanding.
A unified open-vocabulary detector pre-trained on diverse large-scale datasets with language-aware selective fusion, improving category generalization through visual-language alignment.
A temporal memory attention method for long-range video semantic segmentation that captures cross-frame relations without optical-flow prediction.
A framework that fuses OCR and ASR information using a Transformer model with mask/crop strategies and multi-level identity embeddings to generate comprehensive video subtitles.
Any segmentation in images and videos with a single unified MLLM.
From segment anything to any segmentation.
Unified open-vocabulary detection with language-aware selective fusion.
Temporal memory attention for video semantic segmentation.
Systematic Toolchain for AI Research.
Systematic Toolchain for Organizing Research over Years.
Systematic Toolchain for Authoring, Guiding, and Editing.
Oh My DeepSeek Harness plugins: a curated collection of plugins for the DeepSeek Harness.
Template paper for arXiv or any conference.
Run Codex on a remote server.
Run Claude on multiple AI providers.
Developing unified any-segmentation MLLMs for images and videos with Limeng Qiao, Lin Ma and Guanglu Wan.
Co-advised by Prof. Xiaodan Liang and Assoc. Prof. Xiangyuan Lan.
Advised by Prof. Jing Liu.