Tone

All the supervised optical flow networks have this problem, since they are trained on synthesized datasets which have large domain gap with real scenes. I suggest you to train FastFlowNet...

I use TensorRT 6.0 to test FastFlowNet on the Jeston TX2. I write CUDA kernel and TensorRT plugin of Center Dense Dilated Correlation layer and Warping layer, then I use...

Thank you for your appreciation. Let height and width of original input image to be H and W, flow6_up has spatial shape of (H / 32, W / 32), while...

For each decoder, you can change the output channels from 2 to 3, with one channel to estimate occlusion. Also, for the input of each decoder, you should add the...

Please note that each decoder of FastFlowNet estimates optical flow whose magnitude is 1/20 of ground truth flow, therefore, to convert to true displacement before warping, we should multiply a...

Which version of PyTorch do you use? Have you successfully installed the [Pytorch-Correlation-extension](https://github.com/ClementPinard/Pytorch-Correlation-extension) module?

I replace the correlation module with that in FlowNet2 [https://github.com/NVIDIA/flownet2-pytorch](url)

确保你训练测试时embt都是对应正确的

建议先使用train_gopro.py在GoPro数据集上训练并测试，不要用train_vimeo90k.py训练的模型在demo_8x.py脚本中进行测试。

You can search this error and get the solution, such as this [issue discussion](https://discuss.pytorch.org/t/missing-keys-unexpected-keys-in-state-dict-when-loading-self-trained-model/22379/16). Also, I have fixed this problem by changing ddp_model.state_dict() to ddp_model.module.state_dict() when saving model checkpoints.