How to calculate mean Average Recall (mAR)?

[WillianaLeite]

I customized the "https://github.com/matterport/Mask_RCNN.git" repository to train with my own data set, for object detection, ignoring the segmentation part of the mask. Now, I am evaluating my results, I can calculate the MAP through the function: compute_ap, from "https://github.com/matterport/Mask_RCNN/blob/master/mrcnn/utils.py" which returns the "precisions, recalls, overlaps "for each image. I would like to know how can I calculate the mean Average Recall (mAR) using this function: compute_recall from "https://github.com/matterport/Mask_RCNN/blob/master/mrcnn/utils.py"?

[gchai123]

I customized the "https://github.com/matterport/Mask_RCNN.git" repository to train with my own data set, for object detection, ignoring the segmentation part of the mask. Now, I am evaluating my results, I can calculate the MAP through the function: compute_ap, from "https://github.com/matterport/Mask_RCNN/blob/master/mrcnn/utils.py" which returns the "precisions, recalls, overlaps "for each image. I would like to know how can I calculate the mean Average Recall (mAR) using this function: compute_recall from "https://github.com/matterport/Mask_RCNN/blob/master/mrcnn/utils.py"?

Could you please share how did you do the compute_ap()? I cannot figure out where I can get teh gt_boxes, gt_class)ids, gt_masks...........!

[burhr2]

Hi, would also appreciate seeing an answer for @gchai123 question as I am also looking for a way to evaluate my custom test set for segmentation

[Shakesbeer333]

any new insights?

[WillianaLeite]

Hi guys!
I believe I found the solution to this issue. It is available here: issues 2513