picked_idx = pred_conf_val[:,1] > 0.5

[BharathRajM]

The confidence values are in negatives(-1.5,-1.76~), even while evaluating your trained model. Could you recheck this in line #128 of the evaluate.py?

[lyl1140292193]

I encountered the same problem.While evaluating the h5 files in h5_dataset/bunny/cycle_0000.zip with the given trained model,the detected confidence values are all near to zero.Anyway,the selected data with threshold 0.5 is none.

[lvwj19]

Thanks for your attention! We have fixed this problem，and you can try it again.

without confidence filter

with confidence filter