Create run_grasp_planning.py and run_grasp_motion_planning.py, create nerf_input, improve union usage, create frames - Githubissues

tylerlum / get_a_grip

Official implementation of Get a Grip: Multi-Finger Grasp Evaluation at Scale Enables Robust Sim-to-Real Transfer

https://sites.google.com/view/get-a-grip-dataset

MIT License

13 stars 0 forks source link

Create run_grasp_planning.py and run_grasp_motion_planning.py, create nerf_input, improve union usage, create frames #8

Closed tylerlum closed 4 months ago

tylerlum commented 4 months ago

Changes:

Remove old run_ code and _utils code for grasp planning and motion planning
Create run_grasp_planning.py and run_grasp_motion_planning.py
Create nerf_input shared abstraction for interacting with nerf
Improve union usage for tyro
Create frames to handle X_A_B, visualize
Clean up into sampler, evaluator, optimizer
Fix up nerf alphas global computation
Consolidate global grid code
Rename pointcloud to point_cloud
Clean up nerfdata gen
Clean up use of train nerf and generate point cloud
Fix up point cloud export
Cache values repeated for nerf sampler
Fix up use of get nerf densities global