KPConv-PyTorch
KPConv-PyTorch copied to clipboard
During the test, the result of the fourth category is always 0
Confusion on sub clouds 66.03 | 92.96 98.48 82.77 0.00 20.84 55.93 66.44 90.48 80.73 75.29 67.02 69.37 58.07
This is due to the dataset. The beams in Area_5 do not look like the beams in the other areas, and therefore, are never classified as such.
If you use another cloud than Area_5 as validation, you will see this score increase