You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi,
I installed 1.0.0.rc2 and run the dataset generation for the KITTI dataset.
I evaluated the provided pre-trained model for PointRCNN over the KITTI dataset and it doesn't show the performance reported on the paper. The results are:
----------- AP11 Results ------------
Car AP11@0.70, 0.70, 0.70:
bbox AP11:0.0000, 0.0000, 0.0339
bev AP11:0.0000, 0.0866, 0.0866
3d AP11:0.0000, 0.0000, 0.0000
aos AP11:0.00, 0.00, 0.00
Car AP11@0.70, 0.50, 0.50:
bbox AP11:0.0000, 0.0000, 0.0339
bev AP11:0.0000, 0.0866, 0.0866
3d AP11:0.0000, 0.0000, 0.0000
aos AP11:0.00, 0.00, 0.00
Pedestrian AP11@0.50, 0.50, 0.50:
bbox AP11:6.6845, 6.2422, 6.1098
bev AP11:12.7065, 12.0848, 11.5359
3d AP11:0.0000, 0.0000, 0.0019
aos AP11:6.58, 6.15, 5.98
Pedestrian AP11@0.50, 0.25, 0.25:
bbox AP11:6.6845, 6.2422, 6.1098
bev AP11:42.9215, 39.5759, 34.6241
3d AP11:35.1524, 32.6993, 28.4828
aos AP11:6.58, 6.15, 5.98
Cyclist AP11@0.50, 0.50, 0.50:
bbox AP11:3.7611, 3.0747, 2.5931
bev AP11:0.5722, 1.5152, 1.5152
3d AP11:0.0000, 0.0000, 0.0000
aos AP11:3.76, 3.07, 2.59
Cyclist AP11@0.50, 0.25, 0.25:
bbox AP11:3.7611, 3.0747, 2.5931
bev AP11:48.6239, 43.3577, 40.6218
3d AP11:23.8051, 19.6509, 19.2659
aos AP11:3.76, 3.07, 2.59
Overall AP11@easy, moderate, hard:
bbox AP11:3.4818, 3.1056, 2.9123
bev AP11:4.4262, 4.5622, 4.3792
3d AP11:0.0000, 0.0000, 0.0006
aos AP11:3.45, 3.07, 2.86
----------- AP40 Results ------------
Car AP40@0.70, 0.70, 0.70:
bbox AP40:0.0000, 0.0000, 0.0000
bev AP40:0.0000, 0.0000, 0.0000
3d AP40:0.0000, 0.0000, 0.0000
aos AP40:0.00, 0.00, 0.00
Car AP40@0.70, 0.50, 0.50:
bbox AP40:0.0000, 0.0000, 0.0000
bev AP40:0.0000, 0.0198, 0.0198
3d AP40:0.0000, 0.0000, 0.0000
aos AP40:0.00, 0.00, 0.00
Pedestrian AP40@0.50, 0.50, 0.50:
bbox AP40:2.8530, 2.6450, 2.2946
bev AP40:5.2249, 4.6793, 3.7824
3d AP40:0.0000, 0.0000, 0.0000
aos AP40:2.74, 2.53, 2.18
Pedestrian AP40@0.50, 0.25, 0.25:
bbox AP40:2.8530, 2.6450, 2.2946
bev AP40:40.2142, 36.3923, 31.3368
3d AP40:31.1603, 28.6751, 24.0590
aos AP40:2.74, 2.53, 2.18
Cyclist AP40@0.50, 0.50, 0.50:
bbox AP40:3.2933, 2.2465, 2.1000
bev AP40:0.2561, 0.1185, 0.1182
3d AP40:0.0000, 0.0000, 0.0000
aos AP40:3.29, 2.24, 2.10
Cyclist AP40@0.50, 0.25, 0.25:
bbox AP40:3.2933, 2.2465, 2.1000
bev AP40:46.9033, 40.4245, 38.5530
3d AP40:18.5216, 14.2241, 13.6172
aos AP40:3.29, 2.24, 2.10
Overall AP40@easy, moderate, hard:
bbox AP40:2.0488, 1.6305, 1.4649
bev AP40:1.8270, 1.5993, 1.3002
3d AP40:0.0000, 0.0000, 0.0000
aos AP40:2.01, 1.59, 1.42
I also trained the model from scratch with the default config file and couldn't get the reported performance.
I assume it might be related to the refactoring of coordinate systems.
The text was updated successfully, but these errors were encountered:
It should not be related to the refactoring of coordinate systems because PointRCNN is merged after the release of 1.0.0rc.
"couldn't get the reported performance" with training from scratch means similar performance with the one shown above? We can first try to reproduce your problem with the given pretrained model.
Yes. Even if I train the default config file over 8 GPUs, the reported mAP is similar to what I reported above. I tried it with the dataset generated with different rc0, rc1, and 0.18.1. I was not able to get the reported mAP? You're right it may not be related to the coordinate systems.
Hi,
I installed 1.0.0.rc2 and run the dataset generation for the KITTI dataset.
I evaluated the provided pre-trained model for PointRCNN over the KITTI dataset and it doesn't show the performance reported on the paper. The results are:
I also trained the model from scratch with the default config file and couldn't get the reported performance.
I assume it might be related to the refactoring of coordinate systems.
The text was updated successfully, but these errors were encountered: