telmop / tf_flownet2 Goto Github PK

This repo includes the code for running FlowNet 2 on Tensorflow. It does NOT support training. The goal of this code is just to be able to quickly run and use FlowNet 2 using pure Tensorflow, WITHOUT any custom ops. The code here is derived from this repo, which in turn is derived from this. It can deal with inputs of arbitrary sizes.

The main differences between the code here and the one in the repos above are:

• No dependencies on custom ops. All the code here is pure Python code + Tensorflow code. Just install the requirements, download the checkpoints and you should be good to go.
• No training code. The code in this repo is only meant for running the model. If you are interested in training a FlowNet 2 model, it should be relatively simple to add the training code from the repos above back.

This code has been tested with Python 3.7.5 and TensorFlow 2.1.0.

Clone the repo, install the requirements, and download the weights:

git clone https://github.com/vt-vl-lab/tf_flownet2.git
cd tf_flownet2
pip install -r requirements.txt
checkpoints/download.sh

Running the model should be easy, just check the available demo:

python demo.py

[1] E. Ilg, N. Mayer, T. Saikia, M. Keuper, A. Dosovitskiy, T. Brox FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks, IEEE Conference in Computer Vision and Pattern Recognition (CVPR), 2017.

As noted above, this code is based on vt-vl-lab/tf_flownet2 and sampepose/flownet2-tf.