I download the Pretrained actionness classifier on THUMOS14,and run the binary_test.py to get the action score. Then run the gen_bottom_up_proposals.py to get the result, as follows:
I noticed that #15 ,so change the IOU_thresh to [0.5,0.95,0.05] on line170
average # of proposals: 212.375
.....
Average Recall: 31.4444 55.3948
The AR in the paper is 48.9, I have no idea which one of the above result,or have another convention methods to eval ??? and the above result is well or bad ???
I download the Pretrained actionness classifier on THUMOS14,and run the
binary_test.py
to get the action score. Then run thegen_bottom_up_proposals.py
to get the result, as follows: I noticed that #15 ,so change the IOU_thresh to [0.5,0.95,0.05] on line170The AR in the paper is 48.9, I have no idea which one of the above result,or have another convention methods to eval ??? and the above result is well or bad ???