You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The reid checkpoint loss keyworks "meta". so when I used model with reid checkpoint, it is wrong. For example:
$ python demo/demo_mot.py configs/mot/deepsort/deepsort_faster-rcnn_fpn_4e_mot17-private-half.py --input demo/demo.mp4 --output mot.mp4
2021-01-06 19:52:33,660 - mmtrack - INFO - load detector from: checkpoints/fasterrcnn_r50_fpn_4e_mot17-half-64ee2ed4.pth
2021-01-06 19:52:33,861 - mmtrack - INFO - load reid from:checkpoints/reid/tracktor_reid_r50_iter25245-a452f51f.pth
Traceback (most recent call last):
File "demo/demo_mot.py", line 88, in <module>
main()
File "demo/demo_mot.py", line 57, in main
model = init_model(args.config, args.checkpoint, device=args.device)
File "/home/smile/mmtracking/mmtrack/apis/inference.py", line 34, in init_model
model = build_model(config.model)
File "/home/smile/mmtracking/mmtrack/models/builder.py", line 69, in build_model
return build(cfg, MODELS)
File "/home/smile/mmtracking/mmtrack/models/builder.py", line 34, in build
return build_from_cfg(cfg, registry, default_args)
File "/home/smile/tools/mmcv/mmcv/utils/registry.py", line 171, in build_from_cfg
return obj_cls(**args)
File "/home/smile/mmtracking/mmtrack/models/mot/deep_sort.py", line 35, in __init__
self.init_weights(pretrains)
File "/home/smile/mmtracking/mmtrack/models/mot/deep_sort.py", line 49, in init_weights
self.init_module('reid', pretrain['reid'])
File "/home/smile/mmtracking/mmtrack/models/mot/base.py", line 35, in init_module
if 'CLASSES' in checkpoint['meta']:
KeyError: 'meta'
And:
>>> model = torch.load("checkpoints/reid/tracktor_reid_r50_iter25245-a452f51f.pth")
>>> model["meta"]
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
KeyError: 'meta'
So I think the reid checkpoint you release is broken.
The text was updated successfully, but these errors were encountered:
The reid checkpoint loss keyworks "meta". so when I used model with reid checkpoint, it is wrong. For example:
And:
So I think the reid checkpoint you release is broken.
The text was updated successfully, but these errors were encountered: