Training-free framework that converts SAM3 into a real-time multi-class open-vocabulary detector. Achieves 55.8 AP on COCO val2017 (80 classes) at 15.8 FPS (4 classes, 1008px) on a single RTX 4080.

Models can now be quickly used without using this repository using the following code. This can be set up also in other repositories. It has not been widely tested, so we welcome any bug reports / ...