Research
WeChat releases WeVisOne-VL, a vision-language model unifying object and pixel-level perception
The model handles detection, referring localization, counting, OCR, segmentation and matting over 30-plus benchmarks, with a lightweight pixel decoder added via a special token. It is released to the open-source community.
Every outlet that ran it 1 article
Hugging Face PapersResearchLead
WeVisOne-VL: Unified Visual Perception from Objects to Pixels