- Follow the installation steps in the chvmp/champ repository. We've already included customizations for Spot, so no need to follow the additional instructions linked underneath section 1.1.
- Follow the installation steps in the ngp_pl repository. We ran our program under pytorch 1.13.1+cu117.
Navigate to src/
mkdir ../../spot_data
python main.py
Hit ctrl-c after a desired amount of keyframes have been captured, and the program will store the data in the spot_data folder
Navigate to spot-sdk/python/examples/wasd
python wasd.py $ROBOT_IP
In the grey Spot client
- Press 'p' to begin RRT* planning from current location to specified location in wasd.py
- Press number keys 1~9 to navigate and plan to other preset locations instead
- Press 'm' to execute a single step in the plan
- Press 'M' to execute all steps in the plan
- Press 'B' to move arm into position for data collection
- Press 'o' to open the arm (this allows full, unobstructed view for the camera)
- Press 'c' to log the current position in the console
Navigate to NeRF/ngp_pl
python show_gui.py --root_dir ../../../spot_data/ --exp_name Spot --dataset_name spot_vis --ckpt_path ../../../spot_data/ckpts/spot_online/Spot/2_slim.ckpt