search

joslefaure / HIT

Official Implementation of our WACV2023 paper: “Holistic Interaction Transformer Network for Action Detection”
https://arxiv.org/abs/2210.12686
55 stars 9 forks source link

get frame mAP error #34

Open ZWXCV opened 10 months ago

ZWXCV commented 10 months ago

Thanks for the author's reply, I can indeed get videomAP through the instruction. I want to get the frame mAP and I see that the parameter "frameAP " exists in compute_video_map.py, so I execute the instruction "python compute_video_map.py frameAP JHMDB-GT.pkl result_jhmdb.pkl". But I get a syntax error. The following is the error message: Traceback (most recent call last): File "compute_video_map.py", line 427, in frameAP(args.groundtruth, args.detections, args.thr) File "compute_video_map.py", line 86, in frameAP detections = alldets[alldets[:, 2] == ilabel, :] TypeError: unhashable type: 'slice' According to the error message TypeError: unhashable type: 'slice', the slicing operation cannot be applied to NumPy arrays. This makes me very confused, I hope the author can give an answer,thank you very much.

joslefaure commented 10 months ago

Please follow these instructions for frameAP calculations

joslefaure commented 10 months ago

python evaluation_ucf24_jhmdb/pascalvoc.py --gtfolder groundtruths_jhmdb --detfolder PATH-TO-DETECTIONS-FOLDER

ZWXCV commented 10 months ago

Thank you for the author's reply, but why is the value of video mAP that I received significantly different from the value given in your paper, about 5%? I hope the author can provide some suggestions. Thank you very much