Calculate mAP for different scenario ( easy, moderate, hard )

[Soodep]

How do I display the mAP table for different conditions such as easy, moderate and hard. Currently the mAP is calculated and displayed once for each catagory but I want to display the mAP result based on condition such as easy moderate and hard. I am using KITTI dataset and testing results using Faster RCNN.

[ZwwWayne]

The easy, moderate, hard conditions are not implemented. To do that, you need to implement a new function in dataset.evaluate to separate the boxes of the three conditions.

[Soodep]

I would really appreciate if there is any reference or example that is available for such implementation. Please attach the link if there are any resources that are helpful

[shinya7y]

The official "development kit":

• http://www.cvlibs.net/datasets/kitti/eval_object.php?obj_benchmark=2d

Related files in MMDetection3D:

• https://github.com/open-mmlab/mmdetection3d/blob/b4240bd665686b062e987a61ff33c4d5a57242cd/mmdet3d/datasets/kitti_dataset.py#L480
• https://github.com/open-mmlab/mmdetection3d/blob/master/mmdet3d/core/evaluation/kitti_utils/eval.py

[Petros626]

@shinya7y was this comment for the already implemented code or where @Soodep has to change the code?