SenseTime open-sources SenseNova-Vision, a unified vision model for four major tasks

SenseTime launched and fully open-sourced SenseNova-Vision, a unified vision model that integrates understanding and generation. Instead of packaging multiple expert models, it makes vision a native ability of the AI system.

The model unifies four major vision tasks: object detection, image segmentation, depth prediction, and 3D reconstruction. In benchmarks, its single-model performance matches or beats specialized models in structured understanding, dense geometry, segmentation, and multi-view 3D tasks.

SenseTime also open-sourced a 50-million-sample visual instruction dataset, SenseNova-Vision Corpus-50M. The model's core tech will be integrated into the upcoming SenseNova U-series models.