Learning Smoke Removal from Simulation
This repo contains the implementation of CNN based smoke removal from simulation.
If this code is useful for your project, please consider to cite:
Chen, L., Wen, T., John, N.W
Unsupervised Learning of Surgical Smoke Removal from Simulation.
Hamlyn Symposium on Medical Robotics. 2018.
Requirements
This code was tested with Blender, Tensorflow, CUDA 8.0 and Ubuntu 16.04.
Data Preparation
We use the da Vinci dataset from Hamlyn Centre Laparoscopic / Endoscopic Video Datasets
render_smoke.py contains the python script to render smoke on images
Usage:
- Modify the data path in render_smoke.py
- ./blender --python ./render_smoke.py
Training
desmoke_main.py
Testing
desmoke_test.py