This repository contains the demo for dense object nets for the course - Deep Learning for Robotics, Spring'23 at Georgia Tech.
- Clone the git repository:
git clone https://github.com/RobotLocomotion/pytorch-dense-correspondence.git
- Copy the Python3 version of the download script to
pytorch-dense-correspondence
repo.
cp scripts/download_pdc_data.py <path to repo>/config/download_pdc_data.py
- Download the data
cd <path to repo>
python config/download_pdc_data.py ./config/dense_correspondence/dataset/composite/shoes_all.yaml /content/pytorch-dense-correspondence/
- Download model checkpoint and unzip
wget https://data.csail.mit.edu/labelfusion/pdccompressed/trained_models/stable/shoes_consistent_M_background_0.500_3.tar.gz
tar -xvf shoes_consistent_M_background_0.500_3.tar.gz
rm shoes_consistent_M_background_0.500_3.tar.gz