shahabty / EDL-stereo

An unofficial implementation of "Efficient Deep Learning for Stereo Matching"

An unofficial implementation of "Efficient Deep Learning for Stereo Matching"

Getting Started

This implementation is based on Tensorflow 1.12 + tf.keras and Eager Execution

Prerequisites

Tensorflow 1.12 ,Pillow

Installing

1. Create a Anaconda3 environment with pip
2. Activate the environment
3. install tensorflow 1.12
4. Install pillow by running the following command

username@PC:~$ conda install -c anaconda pillow 

5. Go to this link and download "saved_data".
6. Dump it to the project directory.

Training and Validating

1. Training losses and validating errors are saved in "saved_data/runs". Please follow the following instruction to see them in tensorboard.

username@PC:~$ cd saved_dir
username@PC:~$ tensorboard --logdir="runs" --port=6006

Then, open your browser and go to the link below.

username@PC:~$ http://computer_name_or_ip_address:6006

2. Qualitative examples are stored in "saved_data/qualitative".
3. Checkpoints are stored in "saved_data/checkpoints"

Acknowledgments

Thanks to all available implementation of this paper.