Can't get attribute 'SPPF' on <module 'models.common' from '/content/zero-shot-object-tracking/models/common.py

[VYRION-Ai]

i got this error when i run your colab

 File "clip_object_tracker.py", line 360, in <module>
    detect()
  File "clip_object_tracker.py", line 104, in detect
    yolov5_engine = Yolov5Engine(opt.weights, device, opt.classes, opt.confidence, opt.overlap, opt.agnostic_nms, opt.augment, half)
  File "/content/zero-shot-object-tracking/utils/yolov5.py", line 6, in __init__
    self.model = attempt_load(weights, map_location=device)
  File "/content/zero-shot-object-tracking/models/experimental.py", line 118, in attempt_load
    model.append(torch.load(w, map_location=map_location)['model'].float().fuse().eval())  # load FP32 model
  File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 607, in load
    return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
  File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 882, in _load
    result = unpickler.load()
  File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 875, in find_class
    return super().find_class(mod_name, name)
AttributeError: Can't get attribute 'SPPF' on <module 'models.common' from '/content/zero-shot-object-tracking/models/common.py'>

[ikitcheng]

I found this post by programmerah which solved this attribute error.

But now i get another error when I use custom weights:

Namespace(agnostic_nms=False, api_key=None, augment=False, cfg='yolov4.cfg', classes=None, confidence=0.4, detection_engine='yolov5', device='', exist_ok=False, img_size=640, info=False, max_cosine_distance=0.4, name='exp', names='coco.names', nms_max_overlap=1.0, nn_budget=None, overlap=0.3, project='runs/detect', save_conf=False, save_txt=False, source='./data/video/cars.mp4', thickness=3, update=False, url=None, view_img=False, weights=['trial_best.pt'])
Fusing layers... 
Using torch 1.10.0+cu111 CUDA:0 (Tesla K80, 11441.1875MB)

/usr/local/lib/python3.7/dist-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  ../aten/src/ATen/native/TensorShape.cpp:2157.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
Traceback (most recent call last):
  File "clip_object_tracker.py", line 360, in <module>
    detect()
  File "clip_object_tracker.py", line 141, in detect
    _ = yolov5_engine.infer(img.half() if half else img) if device.type != 'cpu' else None  # run once
  File "/content/zero-shot-object-tracking/utils/yolov5.py", line 16, in infer
    pred = self.model(img, augment=self.augment)[0]
  File "/usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "/content/zero-shot-object-tracking/models/yolo.py", line 123, in forward
    return self.forward_once(x, profile)  # single-scale inference, train
  File "/content/zero-shot-object-tracking/models/yolo.py", line 139, in forward_once
    x = m(x)  # run
  File "/usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "/content/zero-shot-object-tracking/models/yolo.py", line 58, in forward
    y[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i]  # wh
RuntimeError: The size of tensor a (80) must match the size of tensor b (84) at non-singleton dimension 3

[ikitcheng]

Hi,

Addressing issues #21 and #15.

So I think updating some of the file content with the latest yolov5 repo will solve the problem.

Please see my branch for a fix.

I have tested it in the following notebook, where I commented out the
!git clone https://github.com/roboflow-ai/zero-shot-object-tracking
and replaced it with my version of the repo instead:
!git clone -b fix-custom-weights https://github.com/ikitcheng/zero-shot-object-tracking.git.

To run zeroshot object tracking with custom weights:
!python clip_object_tracker.py --source ./data/video/cars.mp4 --detection-engine yolov5 --weights custom_weights.pt

Please let me know if it works.

Thanks

[AmartyaCSB]

Hi @ikitcheng

I couldn't see the results being saved at all. Even though the runs are complete, it's a blank folder under exp.
Could you tell me what might be going wrong?

[ikitcheng]

Hi @AmartyaCSB,
What was your final print statement after executing the cell with
!python clip_object_tracker.py --source ./data/video/cars.mp4 --detection-engine yolov7?

Does it print:

Results saved to runs/detect/exp
Done. (XX.XXXs)

This is the notebook.