SenseNova-Vision by SenseTime Open-Sources a Unified Vision Large Model, Achieving Outstanding Performance in Four Core Visual Tasks with a Single Model

SenseTime officially released and fully open-sourced the "SenseNova-Vision" unified visual large model for understanding and generation. This marks a significant visual capability upgrade in SenseTime's SenseNova large model system. SenseTime stated that previous "unified vision" in the industry mostly involved packaging multiple expert models such as detection, segmentation, and depth prediction, which were essentially fragmented. The core innovation of SenseNova-Vision is to make vision a native capability of a general foundation model, thoroughly integrating it into the large model system.

Outperforming Four Key Domains with a Single Model

In various evaluations, SenseNova-Vision leads extensively across four core visual domains as a single model, matching or even surpassing specialized expert models in each domain. In structured visual understanding, tasks such as object detection, referential detection, OCR, and keypoint localization significantly outperform similar general models. Its performance in complex scenarios like dense small-object detection and long-tail category recognition is particularly outstanding. In dense geometric prediction, the accuracy of depth estimation and surface normal estimation reaches the level of dedicated geometric models, maintaining high stability in both indoor and outdoor environments.

Segmentation capabilities cover general segmentation, reasoning segmentation, and interactive segmentation. Thanks to its strong multimodal understanding ability, it performs remarkably well in reasoning segmentation and dialogue-based segmentation. With just a single model, it can complete high-quality multi-view point cloud reconstruction and camera pose estimation, achieving leading performance among general visual approaches.

Completely Outperforming Vision Banana, with Data Open-Sourced Simultaneously

Compared to semantic-oriented models, SenseNova-Vision achieves comprehensive leadership in visual tasks with high requirements on details such as detection, segmentation, and depth. Compared to the generative-oriented model Vision Banana, it demonstrates a comprehensive generational advantage—surpassing and leading in most metrics in authoritative evaluations. Vision Banana can only address two of the four core areas, while SenseNova-Vision covers all tasks including structured understanding, dense geometry, panoramic segmentation, and multi-view 3D.