something wrong in"My_NOCS/data_list/real_val/2/bowl_white_small_norm/list.txt"
Closed this issue · 5 comments
"List.txt" contains the path of "0204_meta.txt". The file (0204_meta.txt)with serial number 1 in the figure has 5 instances, while the GT file(results_real_test_scene_1_0204.pkl) with serial number 2 in the figure has only 4 instances.So this error occurred
0204_meta.txt:
print "results_real_test_scene_1_0204.pkl":
Hi,
The file 0204_meta.txt and results_real_test_scene_1_0204.pkl are provided by the original NOCS dataset. I have checked it, and found that the object can_arizona_tea_norm should not be in this frame, so maybe delete the line 3 4 can_arizona_tea_norm can solve this error.
Maybe this is not a good solution. I encountered the same problem in the files My_NOCS/data/real_val/scene_1/0203_meta.txt, 0204_meta.txt, 0198_meta.txt, 0199_meta.txt, 0200_meta.txt, maybe there are more .
Thanks for mention this, we didn't realize this problem before, it's not something wrong in list.txt but the ground-truth. However, the original NOCS dataset only provide gt_RTs in results_*.pkl, we can only match the data in *_meta.txt and results_*.pkl according to the order. So if we want to solve this, we have to manually check the data one by one, which is a huge work. We will try to find the author of NOCS to solve this later.
Thank you again for report this error! In this case, wrong ground-truth data will be used in benchmark.py, which means the precision may be higher than we report in paper. We may update the precision on NOCS dataset after fix this error.
Okay thank you.