Run Strong Sort Tracker on GPU

[BirenMer]

Search before asking

• [X] I have searched the Yolo Tracking issues and found no similar bug report.

Question

Below is my code for initializing a tracker strong sort or deep_oc_sort, also my problem is no matter what I do the code runs on my CPU insted of my GPU card (I have two GPU's in th system)

---

from pathlib import Path from boxmot import StrongSORT from boxmot import DeepOCSORT

def tracker_init(tracker_name:str='strong_sort',reid_model_path:str='models/osnet_x0_25_msmt17.pt',cuda_device:bool=True,cuda_device_number:int=0): """ Usage: Used to initialize trackers || Default=strong_sort Currently support two Trackers :

1. strong_sort (default)

2. deep_oc_sort Args :

   1. tracker_name(str) -> used to select a traker || Default = strong_sort
   2. reid_model_path(str) -> used to select a reid model || Defualt = osnet_x0_25_msmt17
   3. cuda_device(bool) -> Used to select between CPU and GPU for calculation || Default=GPU
   4. cuda_device_number(int) -> Used to select GPU device (If there exists multiple device) || Default = 0

   Returns : Returns a tracker object

   List of available reid model : ['resnet50', 'resnet101', 'mobilenetv2_x1_0', 'mobilenetv2_x1_4', 'hacnn', 'mlfn', 'osnet_x1_0', 'osnet_x0_75', 'osnet_x0_5', 'osnet_x0_25', 'osnet_ibn_x1_0', 'osnet_ain_x1_0', 'osnet_ain_x0_75', 'osnet_ain_x0_5', 'osnet_ain_x0_25', 'lmbn_n', 'clip'] """ tracker=None if tracker_name=='deep_oc_sort': tracker = DeepOCSORT( model_weights=Path(reid_model_path), # which ReID model to use device=f'cuda:{cuda_device_number}' if cuda_device else 'cpu', fp16=False, )

   else: tracker = StrongSORT( model_weights=Path(reid_model_path), # which ReID model to use device=f'cuda:{cuda_device_number}' if cuda_device else 'cpu', fp16=False, ) return tracker

   ---

Please let me know what I am doing wrong

[mikel-brostrom]

I cannot help you without more information. When CUDA is not available you will receive a message similar to this

ValueError: Invalid CUDA 'device=0' requested. Use 'device=cpu' or pass valid CUDA device(s) if available, i.e. 'device=0' or 'device=0,1,2,3' for Multi-GPU.

torch.cuda.is_available(): False
torch.cuda.device_count(): 0
os.environ['CUDA_VISIBLE_DEVICES']: None
See https://pytorch.org/get-started/locally/ for up-to-date torch install instructions if no CUDA devices are seen by torch.

[BirenMer]

I cannot help you without more information. When CUDA is not available you will receive a message similar to this

ValueError: Invalid CUDA 'device=0' requested. Use 'device=cpu' or pass valid CUDA device(s) if available, i.e. 'device=0' or 'device=0,1,2,3' for Multi-GPU.

torch.cuda.is_available(): False
torch.cuda.device_count(): 0
os.environ['CUDA_VISIBLE_DEVICES']: None
See https://pytorch.org/get-started/locally/ for up-to-date torch install instructions if no CUDA devices are seen by torch.

Sir I am actually getting no error messages, When I see my gpu usage while running the code they are near to zero and my cpu usage is sky rocketing, If I stop the tracking part and run only YOLO for detection then my CPU usage is normal.

So at this point I am also confused why it is happening.

[mikel-brostrom]

Try device=0