Skip to content

test.md

Latest commit

 

History

History
30 lines (21 loc) · 1.42 KB

test.md

File metadata and controls

30 lines (21 loc) · 1.42 KB

Evaluation

GoalFlow first generates trajectories for all test scenarios and then performs testing.
The Goal Point Construction Module is a pluggable component. We provide scores for the predicted goal points, which you can directly use in your own model.

Preparation for Evaluation

Download the following files from Google Drive and place them in the data directory:

  • goalflow_traj_epoch=54-step=18260.ckpt (perception module and trajectory decoder)
  • goal_point_scores.gz (dac scores and distance scores predicted by goal point construction)
  • cluster_points_8192_.npy (goal point vocabulary)
  • goalflow_navi_epoch=99-step=132500.ckpt (optional, perception module and goal point construction module)

Evaluating GoalFlow

  1. First, generate trajectories using run_generate_trajs.sh. The generated trajectories will be stored in your log directory.
  2. Then, run run_goalflow_trajs.sh with your evaluation metric path and the path to the generated trajectories.
sh scripts/generate/run_generate_trajs.sh
sh scripts/evaluation/run_goalflow_trajs.sh

Evaluating the Goal Point Construction Module (Optional)

If needed, you can generate goal_point_scores using run_generate_navi.sh. The generated navigation scores will be stored in your log directory.

sh scripts/generate/run_generate_navi.sh