Using converted onnx model leads to bad results
dalia-jaber opened this issue · comments
I used the script "pytorch2onnx.py" to convert the pytorch models to onnx models. However, the results obtained using the onnx models are different than using the pytorch models for same images. Is this an expected behavior?
had similar problem when i try to convert to onnx, i remember at that time, my problem has something to do with inconsistency of input regarding to RGB/BGR