conda env create -f environment.yml
pip install networkx==2.5 # random fix to make urdfpy work
- Download Blender 2.93.4 for your OS from https://www.blender.org/download/lts/2-93/
- Alias
blender
in your~/.bashrc
(mine looks like the following:)
alias blender="/Applications/Blender.app/Contents/MacOS/Blender"
- After this, run
source ~/.bashrc
- Provide an input trajectory
traj.npy
which isN x 7
and play it back using a Panda URDF in PyBullet
python pybullet_replay.py --joint_angles_file traj.npy
- Render images of the trajectory into a directory called
images
# Note, this is slow on CPU, so try to subsample your trajectories to length 20-30 poses; (TODO: add support for CUDA rendering)
# Note, see the settings at line 249 for different modes of visualizing
blender -b -P render.py
- See the script to configure different camera poses/rendering settings and adding objects to the scene
- Composite multiple frames to create a trajectory visualization (saves to
result.png
)
python overlay_traj.py