A high-level encapsulation for evaluating detection on the Microsoft COCO dataset. It's based on pycocotools.cocoeval.

1. confusion matrix
2. every category metrics (ap ar)
3. summary metrics (map mar)
4. visualization utils

All examples are shown in Test.ipynb

• https://github.com/svpino/tf_object_detection_cm/blob/master/confusion_matrix.py
• cocodataset/cocoapi#282
• https://www.osgeo.cn/matplotlib/gallery/images_contours_and_fields/image_annotated_heatmap.html