search

selflein / GraphNN-Multi-Object-Tracking

Unofficial PyTorch implementation of "Learning a Neural Solver for Multiple Object Tracking"
121 stars 23 forks source link

Test set with YOLOV4 #16

Open MaddaRe opened 2 years ago

MaddaRe commented 2 years ago

Hi @selflein , I'm trying to do testing using a test set on which I've used YOLOV4; the output format is the same as MOT16, so I think I need to do the same as in the "Testing" section without running run_obj_detect.py. Then I directly run preprocessing.py in test mode but the result is empty: for example, I have data/MOT16/test/MOT-xx/gt/gt.txt and the result of preprocessing is output/MOTxx (empty folder). What could my problem be?

selflein commented 2 years ago

Hi, since there is nothing generated, my suspicion is that the loading of the file fails in some way. Make sure that here: https://github.com/selflein/GraphNN-Multi-Object-Tracking/blob/7207e749af02ec8b53d6b1ef75fa16cb0a5b8208/src/data_utils/preprocessing.py#L55 mot_sequences is not empty.

Otherwise I would recommend stepping through the program with a debugger, i.e.,

python -m pdb preprocessing.py ...

to see where it fails.