RuntimeError when run test_tracktor.py

[sfwang20]

Hi Philipp,
Thanks for your excellent work!
I followed README.md to build the environment and when I run test_tracktor.py, it shows the RuntimeError as following:

INFO - main - Initializing object detector(s).
INFO - main - Initializing reID network(s).
INFO - main - Tracking: MOT17-02-FRCNN
0%| | 0/600 [00:02<?, ?it/s]
ERROR - test_tracktor - Failed after 0:00:27!
Traceback (most recent calls WITHOUT Sacred internals):
File "experiments/scripts/test_tracktor.py", line 143, in main
tracker.step(frame_data)
File "d:\research\projects\tracking_wo_bnw\src\tracktor\tracker.py", line 261, in step
self.obj_detect.load_image(blob['img'])
File "d:\research\projects\tracking_wo_bnw\src\tracktor\frcnn_fpn.py", line 75, in load_image
self.features = self.backbone(preprocessed_images.tensors)
File "C:\Users\shinfu\anaconda3\envs\tracking_wo_bnw\lib\site-packages\torch\nn\modules\module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "C:\Users\shinfu\anaconda3\envs\tracking_wo_bnw\lib\site-packages\torchvision\models\detection\backbone_utils.py", line 39, in forward
x = self.body(x)
File "C:\Users\shinfu\anaconda3\envs\tracking_wo_bnw\lib\site-packages\torch\nn\modules\module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "C:\Users\shinfu\anaconda3\envs\tracking_wo_bnw\lib\site-packages\torchvision\models_utils.py", line 63, in forward
x = module(x)
File "C:\Users\shinfu\anaconda3\envs\tracking_wo_bnw\lib\site-packages\torch\nn\modules\module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "C:\Users\shinfu\anaconda3\envs\tracking_wo_bnw\lib\site-packages\torch\nn\modules\conv.py", line 419, in forward
return self._conv_forward(input, self.weight)
File "C:\Users\shinfu\anaconda3\envs\tracking_wo_bnw\lib\site-packages\torch\nn\modules\conv.py", line 416, in _conv_forward
self.padding, self.dilation, self.groups)
RuntimeError: cuDNN error: CUDNN_STATUS_ALLOC_FAILED

I search by google if there are solutions about the error and it seems like the memory of my GPU is not enough.
My GPU is Nvidia GTX 1660Ti with 6 GB memory.
Besides, I also tried change the batch size (to 32) in res50.yml but it didn't work (it seems like testing won't use this file, I'm not sure about that)
Is there any other way for me to get smaller GPU memory usage?
Thanks for your help!
Best regards.

[timmeinhardt]

Yes, the file you changed is just for the training of the reid model. You could run the tracking evaluation on the CPU or reduce the resolution of the images. The latter might hurt the performance though.