Export Custom Model To Onnx

[Throws100]

Hi,

First of all thanks for your great work!

I tried to convert a custom-trained model into an onnx file. Unfortunately, I did not find specific instructions for this. Therefore I tried to rework your example on a yolomodel. (Found here: notebooks/export-onnx-inference-onnxruntime.ipynb)

I managed to work out how to load my custom model, yet it runs into errors:

`def export_yolo_to_onnx(path_model, onnx_output_path):
# Load the YOLO model using torch.hub
model = torch.hub.load('ultralytics/yolov5', 'custom', path=path_model) # Use the local model
model.eval() # Set model to evaluation mode

# Parameters for ONNX export
size = (640, 640)# Resize image to this size for export
size_divisible=   32            # Ensure the size is divisible by 32
score_thresh=   0.25          # Score threshold for NMS
nms_thresh=  0.45            # NMS threshold
opset_version=  12             # ONNX opset version
batch_size=  1                 # Batch size for export

# Export the model to ONNX format with NMS enabled
export_onnx(
    onnx_path=onnx_output_path,
    model=model,
    size=size,
    size_divisible=size_divisible,
    score_thresh=score_thresh,
    nms_thresh=nms_thresh,
    opset_version=opset_version,
    batch_size=batch_size,
)

print(f"Model successfully exported to ONNX at: {onnx_output_path}")`

TypeError Traceback (most recent call last)
in <cell line: 40>()
38
39 # Run the export function
---> 40 export_yolo_to_onnx(path_model, onnx_output_path)

19 frames
/content/yolov5/models/common.py in forward(self, ims, size, augment, profile)
878 files.append(Path(f).with_suffix(".jpg").name)
879 if im.shape[0] < 5: # image in CHW
--> 880 im = im.transpose((1, 2, 0)) # reverse dataloader .transpose(2, 0, 1)
881 im = im[..., :3] if im.ndim == 3 else cv2.cvtColor(im, cv2.COLOR_GRAY2BGR) # enforce 3ch input
882 s = im.shape[:2] # HWC

TypeError: transpose() received an invalid combination of arguments - got (tuple), but expected one of:

• (int dim0, int dim1)
• (name dim0, name dim1)

Maybe you can advise and as soon as it works I can supply you with the notebook and you could add it to the documentation:)

Thanks!