trained model for lidar point pillars

edit

Hi, all

How can I train and export my own model?

I tried https://github.com/k0suke-murakami/tr...

torch.onnx.export(net, example1, "pp.onnx", verbose=False, input_names = input_names) sp_f = torch.ones([1, 64, 496, 432],dtype=torch.float32, device=pillar_x.device ) torch.onnx.export(net.rpn, sp_f, "rpn.onnx", verbose=False) But the onnx model has different size from offered kitti_pretrained_model.

Note that I did not change any parameters from your repository.

How can I properly export it…? Anyone has similar experiences?

Thank you in advance