Dmitry Kurtaev

So, I tried the following steps: ``` git clone https://github.com/dkurt/ssd.pytorch --branch opencv_support --depth 1 cd ssd.pytorch wget https://s3.amazonaws.com/amdegroot-models/ssd300_mAP_77.43_v2.pth python3 export_to_onnx.py --model ssd300_mAP_77.43_v2.pth ``` with `torch==1.4.0` and `onnx==1.6.0` It reports some...

@abuvaneswari, Proposed solution won't work with ONNXRuntime. It's only for OpenCV and OpenVINO. Both of them can import ONNX model directly.

@pichuan, thank you! I also seen this warning locally but it just about TensorFlow version. If the conversion finished with `[SUCCESS]` - everything is fine.

Hi, @Rakshith2597! Can you please check if I try to reproduce model conversion correctly? ```python net = U_Net(1, 2, conv_type='conv_2d', tsm=True, learn=True) net.eval() dummy_inp = torch.randn([1, 1, 128, 160, 160])...

Thanks, @petebankhead! I'll definitely try it once new release will be published.

There is no TensorRT in OpenVINO but there are [ARM](https://github.com/openvinotoolkit/openvino_contrib/tree/master/modules/arm_plugin) and [CUDA](https://github.com/openvinotoolkit/openvino_contrib/tree/master/modules/nvidia_plugin) plugins.

older issue: https://github.com/opencv/opencv/issues/9096

@homm, thanks! Let me check again

@homm, indeed, there is a bug with even sizes. Odd sizes seem produce correct results. For example, using an image from the issue with extra padded zeros: ```python import numpy...

Can this issue be closed? The story about resize compatibility across the libraries is pretty old. However I believe that we have finished with the described problem by this code:...