search

xukechun / Efficient_goal-oriented_push-grasping_synergy

[RAL & IROS 2021] Efficient learning of goal-oriented push-grasping synergy in clutter
60 stars 10 forks source link

Goal object number in test #7

Open littlefiveRobot opened 2 years ago

littlefiveRobot commented 2 years ago

hello , I had trained your model that you given but when I test ,I was not satisfied with the result. I want to know the goal_obj_idx in your test00-09 respectively becase different goal_obj_idx caused different result.

xukechun commented 2 years ago

Hello!

We provide the object number in our ReadMe as follows:

case index-goal object index 0-4, 1-7, 2-8, 3-7, 4-7, 5-3, 6-1, 7-7, 8-5, 9-9

littlefiveRobot commented 2 years ago

Thank you very much！ I still want to ask a question. When I train the grasp model in stage three, I find the grasp rate as follows, is this normal?(I use the plot.py in VPG) 2022-04-07 09-37-47 的屏幕截图

littlefiveRobot commented 2 years ago

My command: python3 main.py --stage push_only --alternating_training --grasp_goal_conditioned --goal_conditioned --goal_obj_idx 1 --experience_replay --explore_rate_decay --save_visualizations --heuristic_bootstrap --load_snapshot --snapshot_file 'model.pth' The model.pth is provided by you. And I expect the grasp success rate to be greater than 90% .