load model by torch.hub.load() and export polys pixel

[quangdungluong]

Load model by torch.hub.load() can be used to export polys bounding box.
Example:

model = torch.hub.load()
img = "path/to/image"
pred = model(img)
print(pred.polys)

[Wu-Jian-Ting]

Hi, I use this and use Realsense D435i, but it had something error

  File "main.py", line 89, in <module>
    results = model(color_image)
  File "/home/iris/anaconda3/envs/yolov5_obb/lib/python3.8/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/home/iris/anaconda3/envs/yolov5_obb/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 26, in decorate_context
    return func(*args, **kwargs)
  File "/home/iris/yolov5_obb/models/common.py", line 531, in forward
    return Detections(imgs, pred, files, t, self.names, x.shape)
  File "/home/iris/yolov5_obb/models/common.py", line 548, in __init__
    self.xyxy = [poly2hbb(p[:, :8]) for p in pred]
  File "/home/iris/yolov5_obb/models/common.py", line 548, in <listcomp>
    self.xyxy = [poly2hbb(p[:, :8]) for p in pred]
  File "/home/iris/yolov5_obb/utils/rboxs_utils.py", line 156, in poly2hbb
    assert polys.shape[-1] == 8
AssertionError

I print this y = non_max_suppression_obb(y, self.conf, self.iou, self.classes, self.agnostic, self.multi_label, max_det=self.max_det) and I got y: [tensor([], device='cuda:0', size=(0, 7))]

Do somebody konw how to fix it? tks