Step1: train a pcn takes complete points clouds as input

[leonardozcm]

(ML) chriskafka@bigshot:~/PycharmProjects/ASFM-Net-Review$ python main_pcn.py --test --backbone
cuda available  True
Loaded compiled 3D CUDA chamfer distance
Test[1200/1200] Taxonomy = 04530566 Sample = 294644520ccc2ce27795dd28016933fc Losses = ['7.0097', '5.2598', '7.5357'] Metrics = ['0.0053']: 100%|████████████████████████████████████████████████████████████████████████████████| 1200/1200 [02:24<00:00,  8.31it/s]
============================ TEST RESULTS ============================
Taxonomy        #Sample ChamferDistance
02691156        150     0.0056
02933112        150     0.0108
02958343        150     0.0091
03001627        150     0.0102
03636649        150     0.0126
04256520        150     0.0104
04379243        150     0.0099
04530566        150     0.0091
Overall                 0.0097

Epoch  -1       12.8742 9.6951  13.8437

                    4096            16384

Epoch -1 12.8742 9.6951 13.8437

[leonardozcm]

============================ TEST RESULTS ============================
Taxonomy        #Sample ChamferDistance
02691156        150     0.0108
02933112        150     0.0220
02958343        150     0.0184
03001627        150     0.0196
03636649        150     0.0196
04256520        150     0.0202
04379243        150     0.0198
04530566        150     0.0165
Overall                 0.0184

Epoch  -1       17.1826   18.3531              19.0180
                        1024      16384->1024