March 3, 2019

511 words 3 mins read



Pytorch implementation of FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks

repo name NVIDIA/flownet2-pytorch
repo link
language Python
size (curr.) 6433 kB
stars (curr.) 1720
created 2017-11-20
license Other


Pytorch implementation of FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks.

Multiple GPU training is supported, and the code provides examples for training or inference on MPI-Sintel clean and final datasets. The same commands can be used for training or inference with other datasets. See below for more detail.

Inference using fp16 (half-precision) is also supported.

For more help, type

python --help

Network architectures

Below are the different flownet neural network architectures that are provided. A batchnorm version for each network is also available.

  • FlowNet2S
  • FlowNet2C
  • FlowNet2CS
  • FlowNet2CSS
  • FlowNet2SD
  • FlowNet2

Custom layers

FlowNet2 or FlowNet2C* achitectures rely on custom layers Resample2d or Correlation. A pytorch implementation of these layers with cuda kernels are available at ./networks. Note : Currently, half precision kernels are not available for these layers.

Data Loaders

Dataloaders for FlyingChairs, FlyingThings, ChairsSDHom and ImagesFromFolder are available in

Loss Functions

L1 and L2 losses with multi-scale support are available in


# get flownet2-pytorch source
git clone
cd flownet2-pytorch

# install custom layers

Python requirements

Currently, the code supports python 3

  • numpy
  • PyTorch ( == 0.4.1, for <= 0.4.0 see branch python36-PyTorch0.4)
  • scipy
  • scikit-image
  • tensorboardX
  • colorama, tqdm, setproctitle

Converted Caffe Pre-trained Models

We’ve included caffe pre-trained models. Should you use these pre-trained weights, please adhere to the license agreements.


# Example on MPISintel Clean   
python --inference --model FlowNet2 --save_flow --inference_dataset MpiSintelClean \
--inference_dataset_root /path/to/mpi-sintel/clean/dataset \
--resume /path/to/checkpoints 

Training and validation

# Example on MPISintel Final and Clean, with L1Loss on FlowNet2 model
python --batch_size 8 --model FlowNet2 --loss=L1Loss --optimizer=Adam --optimizer_lr=1e-4 \
--training_dataset MpiSintelFinal --training_dataset_root /path/to/mpi-sintel/final/dataset  \
--validation_dataset MpiSintelClean --validation_dataset_root /path/to/mpi-sintel/clean/dataset

# Example on MPISintel Final and Clean, with MultiScale loss on FlowNet2C model 
python --batch_size 8 --model FlowNet2C --optimizer=Adam --optimizer_lr=1e-4 --loss=MultiScale --loss_norm=L1 \
--loss_numScales=5 --loss_startScale=4 --optimizer_lr=1e-4 --crop_size 384 512 \
--training_dataset FlyingChairs --training_dataset_root /path/to/flying-chairs/dataset  \
--validation_dataset MpiSintelClean --validation_dataset_root /path/to/mpi-sintel/clean/dataset

Results on MPI-Sintel

Predicted flows on MPI-Sintel


If you find this implementation useful in your work, please acknowledge it appropriately and cite the paper:

  author       = "E. Ilg and N. Mayer and T. Saikia and M. Keuper and A. Dosovitskiy and T. Brox",
  title        = "FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks",
  booktitle    = "IEEE Conference on Computer Vision and Pattern Recognition (CVPR)",
  month        = "Jul",
  year         = "2017",
  url          = ""
  author = {Fitsum Reda and Robert Pottorff and Jon Barker and Bryan Catanzaro},
  title = {flownet2-pytorch: Pytorch implementation of FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks},
  year = {2017},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{}}

Code (in Caffe and Pytorch): PWC-Net Paper : PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume.


Parts of this code were derived, as noted in the code, from ClementPinard/FlowNetPytorch.

comments powered by Disqus