[SenseTime Open-Sources 7B Vision Model: One Model Handles Detection, Segmentation, and 3D] According to monitoring by Beating, SenseTime has released SenseNova-Vision-7B-MoT. This model unifies tasks such as object detection, OCR, keypoint localization, image segmentation, depth estimation, and multi-view 3D reconstruction into text or image generation tasks. Users can specify their needs using language or visual prompts without having to install dedicated modules for each task. In RefCOCOg referring segmentation and Cityscapes semantic segmentation benchmarks, the model achieved scores of 80.3 and 71.2, respectively, surpassing Google DeepMind Vision Banana's scores of 73.8 and 69.9. SenseTime has simultaneously open-sourced the inference code, model weights, and 50 million visual instruction data points. The code is licensed under Apache 2.0, while the model weights are licensed under CC BY-NC 4.0, restricted to non-commercial use. [Original Link]