CODEWITHSUNDEEP
deep-learningcaffeconvolutiondeconvolutioncaffe2
semenbari
semenbari

Reputation: 805

What does mask r-cnn's AP, AP50, AP70 mean?

I'm novice on r-cnn.

There are term AP, AP50, AP75 on mask r-cnn paper. 50, 75 is small postfix, but I can't make it small, sorry. Anyway the paper says it is averaged over IOU thresholds. For AP50, only candidates over 50% region comparing ground truth are counted, and for AP75 only candidate over 75% are counted. Then what is empty AP? I knew AP use 70% IoU threshold, but it wouldn't. Because just AP is lower then AP75.

And additionally, there are another terms I don't understand well. They are APs, m, L. I know they mean small, medium, large. Then is there any criteria how big they are. Just saying small, medium, large makes me little bit confused.

Upvotes: 11

Views: 10637

Answers (1)

semenbari
semenbari

Reputation: 805

I found them from below: http://cocodataset.org/#detections-eval

Average Precision (AP):

AP         % AP at IoU=.50:.05:.95 **(primary challenge metric)**
APIoU=.50   % AP at IoU=.50 (PASCAL VOC metric)
APIoU=.75   % AP at IoU=.75 (strict metric)

AP Across Scales:

APsmall    % AP for small objects: area < 322
APmedium   % AP for medium objects: 322 < area < 962
APlarge    % AP for large objects: area > 962

Average Recall (AR):

ARmax=1    % AR given 1 detection per image
ARmax=10   % AR given 10 detections per image
ARmax=100  % AR given 100 detections per image

AR Across Scales:

ARsmall    % AR for small objects: area < 322
ARmedium   % AR for medium objects: 322 < area < 962
ARlarge    % AR for large objects: area > 962

Upvotes: 8

Related Questions