votenet-sunrgbd visualization error

[yangninghua]

There is a problem with the visualization, it seems to be related to the rotation angle

mmdetection3d = mmdetection3d-1.0.0rc1

python demo/pcd_demo.py ./data/sunrgbd/points/000015.bin configs/votenet/votenet_16x8_sunrgbd-3d-10class.py ./work_dirs/myvotenet_16x8_sunrgbd-3d-10class/epoch_14.pth --device cuda:1 --score-thr 0.3 --out-dir ../demo_result --show --snapshot

[yangninghua]

open3d 0.15.2 opencv-contrib-python 3.4.3.18 opencv-python 3.4.3.18

[yangninghua]

2022-04-25 17:08:14,658 - mmdet - INFO - Epoch [14][1600/1652]  lr: 8.000e-03, eta: 7:55:24, time: 0.792, data_time: 0.012, memory: 8130, vote_loss: 3.5798, objectness_loss: 0.3804, semantic_loss: 0.4283, center_loss: 0.5644, dir_class_loss: 0.9547, dir_res_loss: 0.7032, size_class_loss: 0.4283, size_res_loss: 0.2672, loss: 7.3063, grad_norm: 7.8146
2022-04-25 17:08:54,242 - mmdet - INFO - Epoch [14][1650/1652]  lr: 8.000e-03, eta: 7:54:45, time: 0.792, data_time: 0.013, memory: 8130, vote_loss: 3.6214, objectness_loss: 0.3990, semantic_loss: 0.4534, center_loss: 0.5275, dir_class_loss: 1.0011, dir_res_loss: 0.6844, size_class_loss: 0.4534, size_res_loss: 0.3061, loss: 7.4464, grad_norm: 8.4020
2022-04-25 17:08:55,907 - mmdet - INFO - Saving checkpoint at 14 epochs
[>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>] 5050/5050, 12.0 task/s, elapsed: 420s, ETA:     0s

2022-04-25 17:16:32,371 - mmdet - INFO - 
+-------------+---------+---------+---------+---------+
| classes     | AP_0.25 | AR_0.25 | AP_0.50 | AR_0.50 |
+-------------+---------+---------+---------+---------+
| bed         | 0.7831  | 0.9184  | 0.3601  | 0.4971  |
| table       | 0.4491  | 0.8130  | 0.0954  | 0.2632  |
| sofa        | 0.4614  | 0.8102  | 0.0952  | 0.2568  |
| chair       | 0.6985  | 0.8595  | 0.2875  | 0.4797  |
| toilet      | 0.4144  | 0.8690  | 0.0143  | 0.1586  |
| desk        | 0.1554  | 0.7561  | 0.0240  | 0.1849  |
| dresser     | 0.1229  | 0.5734  | 0.0071  | 0.1239  |
| night_stand | 0.2153  | 0.6588  | 0.0003  | 0.0353  |
| bookshelf   | 0.0699  | 0.4397  | 0.0006  | 0.0390  |
| bathtub     | 0.4588  | 0.8367  | 0.0733  | 0.2653  |
+-------------+---------+---------+---------+---------+
| Overall     | 0.3829  | 0.7535  | 0.0958  | 0.2304  |
+-------------+---------+---------+---------+---------+
2022-04-25 17:16:32,811 - mmdet - INFO - Exp name: myvotenet_16x8_sunrgbd-3d-10class.py
2022-04-25 17:16:32,812 - mmdet - INFO - Epoch(val) [14][5050]  bed_AP_0.25: 0.7831, table_AP_0.25: 0.4491, sofa_AP_0.25: 0.4614, chair_AP_0.25: 0.6985, toilet_AP_0.25: 0.4144, desk_AP_0.25: 0.1554, dresser_AP_0.25: 0.1229, night_stand_AP_0.25: 0.2153, bookshelf_AP_0.25: 0.0699, bathtub_AP_0.25: 0.4588, mAP_0.25: 0.3829, bed_rec_0.25: 0.9184, table_rec_0.25: 0.8130, sofa_rec_0.25: 0.8102, chair_rec_0.25: 0.8595, toilet_rec_0.25: 0.8690, desk_rec_0.25: 0.7561, dresser_rec_0.25: 0.5734, night_stand_rec_0.25: 0.6588, bookshelf_rec_0.25: 0.4397, bathtub_rec_0.25: 0.8367, mAR_0.25: 0.7535, bed_AP_0.50: 0.3601, table_AP_0.50: 0.0954, sofa_AP_0.50: 0.0952, chair_AP_0.50: 0.2875, toilet_AP_0.50: 0.0143, desk_AP_0.50: 0.0240, dresser_AP_0.50: 0.0071, night_stand_AP_0.50: 0.0003, bookshelf_AP_0.50: 0.0006, bathtub_AP_0.50: 0.0733, mAP_0.50: 0.0958, bed_rec_0.50: 0.4971, table_rec_0.50: 0.2632, sofa_rec_0.50: 0.2568, chair_rec_0.50: 0.4797, toilet_rec_0.50: 0.1586, desk_rec_0.50: 0.1849, dresser_rec_0.50: 0.1239, night_stand_rec_0.50: 0.0353, bookshelf_rec_0.50: 0.0390, bathtub_rec_0.50: 0.2653, mAR_0.50: 0.2304

[Xiangxu-0103]

Whether do you use the latest code? I think this bug has been fixed in PR #1407.

[ZCMax]

yeah, you can update your code, since the visualization bug has been fixed by #1407.