Hello, from the explanation in training guide, I assumed that the box_warp is at least slightly related to extrinsic matrix (maybe I am totally wrong), but it sort of worked since the Shapenet car dataset has extrinsic matrix in range (-1.3,1.3) and box_warp = 1.6, also for another car dataset for which extrinsic matrix in range (-10,10) I set up the parameter bow_warp = 15 and it worked quite well. But I have another dataset which has extrinsic matrix in range (-1020,1020) and there it kind of fails. Also, the training is very slow (at least 5 times slower than in previous cases).
If anyone can give me a hint about the bow_warp, I would be happy :)
Hello, from the explanation in training guide, I assumed that the box_warp is at least slightly related to extrinsic matrix (maybe I am totally wrong), but it sort of worked since the Shapenet car dataset has extrinsic matrix in range (-1.3,1.3) and box_warp = 1.6, also for another car dataset for which extrinsic matrix in range (-10,10) I set up the parameter bow_warp = 15 and it worked quite well. But I have another dataset which has extrinsic matrix in range (-1020,1020) and there it kind of fails. Also, the training is very slow (at least 5 times slower than in previous cases). If anyone can give me a hint about the bow_warp, I would be happy :)