/sde

Stereo Depth Estimation

Primary LanguageJupyter Notebook

Stereo Depth Estimation

This project was done on Ubuntu 18.04 with NVIDIA Titan X (cuda 11.1)

Dependencies

  • conda
  • git
  • cuda
  • data_scene_flow.zip Dataset which can be downloaded from here

Running the project

git clone https://github.com/duklin/sde.git
cd sde

# download TensorBoard Logs and checkpoints
wget --output-document logs_and_checkpoints.tar.gz https://uni-bonn.sciebo.de/s/GShVSpQ7wXsPkRL/download
tar -xvf logs_and_checkpoints.tar.gz

# Create and activate virtual environment
conda env create -n sde -f environment.yml
conda activate sde

# Preprocessing the dataset files
python -m src.dataset.preprocess --root dataset/ --kitti-archive /path/to/data_scene_flow.zip

# Start Jupyter-Lab environment and TensorBoard in separate processes
jupyter-lab --no-browser
tensorboard --logdir logs/ --samples_per_plugin images=30