HF RL Explorer

I want a pure library function bbox.utils.nms(bbl, scores, thresh) for non-maximum suppression over 2D…

I want a pure library function bbox.utils.nms(bbl, scores, thresh) for non-maximum suppression over 2D…: a task in MiMo-V2.6-RL-harbor-code: MiMo-V2.6-RL Code (Harbor) (Harbor dataset). For BBox2DList( 0, 0, 10, 10 , 0, 0, 10, 10 , 20, 20, 5, 5 ) , scores= 0.2, 0.9, 0.5 , and thresh=0.5 , it…

The task

For `BBox2DList([[0, 0, 10, 10], [0, 0, 10, 10], [20, 20, 5, 5]])`, `scores=[0.2, 0.9, 0.5]`, and `thresh=0.5`, it should return `np.array([1, 2], dtype=int)` because the highest-scoring duplicate is kept and the lower-scoring duplicate is suppressed. For `BBox2DList([[0, 0, 10, 10], [5, 5, 10, 10]])`,…

Part of FineEnvs/MiMo-V2.6-RL-harbor-code.