Closed M-o-magic closed 5 months ago
Hi, thanks for your interest. The used coco-style evaluation has the APIs for recall and precision, as shown in pycocotools.
You can use the generated json prediction file and GT json to evaluate with pycocotools. Besides, you can also modify the coco_eval.py here, which may be a little bit complicated. https://github.com/CityU-AIM-Group/SIGMA/blob/d8265ac8ec0e74ef84298f25cae532d4fe67302e/fcos_core/data/datasets/evaluation/coco/coco_eval.py#L52
Hi, thanks for your interest. The used coco-style evaluation has the APIs for recall and precision, as shown in pycocotools.
You can use the generated json prediction file and GT json to evaluate with pycocotools. Besides, you can also modify the coco_eval.py here, which may be a little bit complicated. https://github.com/CityU-AIM-Group/SIGMA/blob/d8265ac8ec0e74ef84298f25cae532d4fe67302e/fcos_core/data/datasets/evaluation/coco/coco_eval.py#L52