2022-10-25 18:47:20,377 adals INFO: 1 GPUs available
2022-10-25 18:47:20,377 adals INFO: Namespace(ckpt2d='@/model_2d_100000.pth', ckpt3d='@/model_3d_065000.pth', config_file='ADALS/config/nuscenes/usa2singapore/adals_sn_sgd.yaml')
2022-10-25 18:47:20,378 adals INFO: Loaded configuration file ADALS/config/nuscenes/usa2singapore/adals_sn_sgd.yaml
2022-10-25 18:47:20,378 adals INFO: Running with config:
AUTO_RESUME: True
DATALOADER:
  DROP_LAST: True
  NUM_WORKERS: 4
DATASET_SOURCE:
  NuScenesSCN:
    augmentation:
      cut_image: True
      cut_range: [512, 512]
      flip_x: 0.5
      height: 32
      noisy_rot: 0.1
      random_flip: True
      remove_large: True
      rot_z: 6.2831
      transl: True
      width: 1024
    full_scale: 4096
    lidarseg: False
    merge_classes: True
    root_dir: /data/datasets/nuScenes/nuscenes_preprocess_lidar/preprocess
    scale: 20
  TRAIN: ('train_usa',)
  TYPE: NuScenesSCN
DATASET_TARGET:
  NuScenesSCN:
    augmentation:
      cut_image: True
      cut_range: [512, 512]
      flip_x: 0.5
      height: 32
      noisy_rot: 0.1
      random_flip: True
      remove_large: True
      rot_z: 6.2831
      transl: True
      width: 1024
    full_scale: 4096
    lidarseg: False
    merge_classes: True
    root_dir: /data/datasets/nuScenes/nuscenes_preprocess_lidar/preprocess
    scale: 20
  TEST: ('test_singapore',)
  TRAIN: ('train_singapore',)
  TYPE: NuScenesSCN
  VAL: ('val_singapore',)
MODEL:
  TYPE: 
MODEL_2D:
  CKPT_PATH: 
  DUAL_HEAD: True
  NUM_CLASSES: 5
  SalsaNextSeg:
    in_channels: 5
  TYPE: SalsaNextSeg
MODEL_3D:
  CKPT_PATH: 
  DUAL_HEAD: True
  NUM_CLASSES: 5
  SCN:
    block_reps: 1
    full_scale: 4096
    in_channels: 1
    m: 16
    num_planes: 7
    residual_blocks: False
  TYPE: SCN
OPTIMIZER:
  BASE_LR: 0.001
  TYPE: 
  WEIGHT_DECAY: 0.0
OPTIMIZER_2D:
  BASE_LR: 0.01
  SGD:
    dampening: 0.0
    momentum: 0.9
  TYPE: SGD
  WEIGHT_DECAY: 0.0
OPTIMIZER_3D:
  Adam:
    betas: (0.9, 0.999)
  BASE_LR: 0.001
  TYPE: Adam
  WEIGHT_DECAY: 0.0
OPTIMIZER_DIS:
  Adam:
    betas: (0.9, 0.99)
  BASE_LR: 0.0001
  TYPE: Adam
  WEIGHT_DECAY: 0.0
OUTPUT_DIR: /workspace/DA-LPS/ADALS/output/@
RESUME_PATH: 
RESUME_STATES: True
RNG_SEED: 1
SCHEDULER:
  CLIP_LR: 0.0
  MAX_ITERATION: 100000
  MultiStepLR:
    gamma: 0.1
    milestones: (80000, 90000)
  TYPE: MultiStepLR
SCHEDULER_DIS:
  PolyLR:
    power: 0.9
  TYPE: PolyLR
TRAIN:
  ADALS:
    lambda_D_src_2d_feat: 0.05
    lambda_D_src_2d_pred: 0.1
    lambda_D_src_3d_pred: 0.1
    lambda_D_trg_2d_feat: 0.05
    lambda_D_trg_2d_pred: 0.2
    lambda_D_trg_3d_pred: 0.2
    lambda_G_trg_2d_feat: 0.001
    lambda_G_trg_2d_pred: 0.07
    lambda_G_trg_3d_pred: 0.05
    lambda_logcoral: 0.0
    lambda_minent: 0.0
    lambda_pixels: 0.5
    lambda_xm_src: 0.8
    lambda_xm_tgl: 0.0
    lambda_xm_trg: 0.1
  BATCH_SIZE: 8
  CHECKPOINT_PERIOD: 5000
  CLASS_WEIGHTS: [2.42607725, 4.61879653, 5.72875704, 3.69461499, 1.0]
  FROZEN_PATTERNS: ()
  LOG_PERIOD: 50
  MAX_TO_KEEP: 100
  SUMMARY_PERIOD: 50
VAL:
  BATCH_SIZE: 8
  KNN:
    cutoff: 1.0
    knn: 5
    search: 5
    sigma: 1.0
  LOG_PERIOD: 20
  METRIC: seg_iou
  PERIOD: 5000
2022-10-25 18:47:26,806 adals.validate INFO: Validation
2022-10-25 18:47:32,989 adals.validate INFO: iter: 1/367  seg_loss_2d_point: 0.0462 (0.0462)  seg_loss_2d_pixel: 0.0744 (0.0744)  seg_loss_3d: 0.0413 (0.0413)  time: 6.1815 (6.1815)  data: 0.6706 (0.6706)  max mem: 2853
2022-10-25 18:48:34,387 adals.validate INFO: iter: 20/367  seg_loss_2d_point: 0.0274 (0.0274)  seg_loss_2d_pixel: 0.0461 (0.0461)  seg_loss_3d: 0.0356 (0.0356)  time: 3.3790 (3.3790)  data: 0.0861 (0.0861)  max mem: 3337
2022-10-25 18:49:31,886 adals.validate INFO: iter: 40/367  seg_loss_2d_point: 0.0638 (0.0456)  seg_loss_2d_pixel: 0.0900 (0.0680)  seg_loss_3d: 0.0681 (0.0519)  time: 2.8749 (3.1270)  data: 0.0634 (0.0748)  max mem: 3650
2022-10-25 18:50:10,423 adals.validate INFO: iter: 60/367  seg_loss_2d_point: 0.0417 (0.0443)  seg_loss_2d_pixel: 0.0558 (0.0639)  seg_loss_3d: 0.0449 (0.0496)  time: 1.9269 (2.7269)  data: 0.0625 (0.0707)  max mem: 3650
2022-10-25 18:51:28,593 adals.validate INFO: iter: 80/367  seg_loss_2d_point: 0.0343 (0.0418)  seg_loss_2d_pixel: 0.0609 (0.0632)  seg_loss_3d: 0.0410 (0.0474)  time: 3.9085 (3.0223)  data: 0.0641 (0.0690)  max mem: 4852
2022-10-25 18:52:24,207 adals.validate INFO: iter: 100/367  seg_loss_2d_point: 0.0317 (0.0398)  seg_loss_2d_pixel: 0.0577 (0.0621)  seg_loss_3d: 0.0322 (0.0444)  time: 2.7807 (2.9740)  data: 0.0603 (0.0673)  max mem: 4852
2022-10-25 18:53:47,606 adals.validate INFO: iter: 120/367  seg_loss_2d_point: 0.0708 (0.0449)  seg_loss_2d_pixel: 0.0986 (0.0682)  seg_loss_3d: 0.0710 (0.0488)  time: 4.1699 (3.1733)  data: 0.0618 (0.0664)  max mem: 4970
2022-10-25 18:55:48,772 adals.validate INFO: iter: 140/367  seg_loss_2d_point: 0.0568 (0.0466)  seg_loss_2d_pixel: 0.0828 (0.0703)  seg_loss_3d: 0.0696 (0.0518)  time: 6.0583 (3.5855)  data: 0.0616 (0.0657)  max mem: 4970
2022-10-25 18:56:45,174 adals.validate INFO: iter: 160/367  seg_loss_2d_point: 0.0400 (0.0458)  seg_loss_2d_pixel: 0.0546 (0.0683)  seg_loss_3d: 0.0375 (0.0500)  time: 2.8201 (3.4898)  data: 0.0649 (0.0656)  max mem: 4970
2022-10-25 18:57:25,895 adals.validate INFO: iter: 180/367  seg_loss_2d_point: 0.0354 (0.0446)  seg_loss_2d_pixel: 0.0561 (0.0669)  seg_loss_3d: 0.0364 (0.0485)  time: 2.0360 (3.3283)  data: 0.0591 (0.0649)  max mem: 4970
2022-10-25 18:58:29,142 adals.validate INFO: iter: 200/367  seg_loss_2d_point: 0.0323 (0.0434)  seg_loss_2d_pixel: 0.0607 (0.0663)  seg_loss_3d: 0.0341 (0.0471)  time: 3.1623 (3.3117)  data: 0.0619 (0.0646)  max mem: 4970
2022-10-25 18:59:19,288 adals.validate INFO: iter: 220/367  seg_loss_2d_point: 0.0308 (0.0423)  seg_loss_2d_pixel: 0.0622 (0.0660)  seg_loss_3d: 0.0356 (0.0460)  time: 2.5073 (3.2385)  data: 0.0633 (0.0645)  max mem: 4970
2022-10-25 19:00:15,245 adals.validate INFO: iter: 240/367  seg_loss_2d_point: 0.0351 (0.0417)  seg_loss_2d_pixel: 0.0619 (0.0656)  seg_loss_3d: 0.0382 (0.0454)  time: 2.7979 (3.2018)  data: 0.0635 (0.0644)  max mem: 4970
2022-10-25 19:01:36,576 adals.validate INFO: iter: 260/367  seg_loss_2d_point: 0.0360 (0.0412)  seg_loss_2d_pixel: 0.0587 (0.0651)  seg_loss_3d: 0.0305 (0.0442)  time: 4.0665 (3.2683)  data: 0.0633 (0.0643)  max mem: 4970
2022-10-25 19:03:07,797 adals.validate INFO: iter: 280/367  seg_loss_2d_point: 0.0598 (0.0426)  seg_loss_2d_pixel: 0.0896 (0.0668)  seg_loss_3d: 0.0630 (0.0456)  time: 4.5610 (3.3607)  data: 0.0593 (0.0639)  max mem: 5079
2022-10-25 19:04:05,269 adals.validate INFO: iter: 300/367  seg_loss_2d_point: 0.0308 (0.0418)  seg_loss_2d_pixel: 0.0628 (0.0666)  seg_loss_3d: 0.0500 (0.0458)  time: 2.8736 (3.3282)  data: 0.0597 (0.0637)  max mem: 5079
2022-10-25 19:05:02,551 adals.validate INFO: iter: 320/367  seg_loss_2d_point: 0.0269 (0.0408)  seg_loss_2d_pixel: 0.0536 (0.0658)  seg_loss_3d: 0.0492 (0.0461)  time: 2.8641 (3.2992)  data: 0.0640 (0.0637)  max mem: 5079
2022-10-25 19:05:59,003 adals.validate INFO: iter: 340/367  seg_loss_2d_point: 0.0300 (0.0402)  seg_loss_2d_pixel: 0.0496 (0.0648)  seg_loss_3d: 0.0577 (0.0467)  time: 2.8226 (3.2712)  data: 0.0610 (0.0635)  max mem: 5079
2022-10-25 19:06:41,791 adals.validate INFO: iter: 360/367  seg_loss_2d_point: 0.0330 (0.0398)  seg_loss_2d_pixel: 0.0522 (0.0641)  seg_loss_3d: 0.0387 (0.0463)  time: 2.1394 (3.2083)  data: 0.0586 (0.0633)  max mem: 5079
2022-10-25 19:07:09,805 adals.validate INFO: 2D Point overall accuracy=98.79%
2022-10-25 19:07:09,806 adals.validate INFO: 2D Point overall IOU=58.19
2022-10-25 19:07:09,813 adals.validate INFO: 2D Point class-wise segmentation accuracy and IoU.
+------------------+------------+-------+----------+
| Class            |   Accuracy |   IOU |    Total |
|------------------+------------+-------+----------|
| vehicle          |      90.49 | 76.07 |  2605498 |
| pedestrian       |      68.85 | 43.55 |   244248 |
| bike             |      24.85 | 17.25 |    92887 |
| traffic_boundary |      73.66 | 55.27 |   388136 |
| background       |      99.25 | 98.82 | 98362495 |
+------------------+------------+-------+----------+
2022-10-25 19:07:09,813 adals.validate INFO: 2D Pixel overall accuracy=98.24%
2022-10-25 19:07:09,813 adals.validate INFO: 2D Pixel overall IOU=55.56
2022-10-25 19:07:09,814 adals.validate INFO: 2D Pixel class-wise segmentation accuracy and IoU.
+------------------+------------+-------+----------+
| Class            |   Accuracy |   IOU |    Total |
|------------------+------------+-------+----------|
| vehicle          |      90.88 | 71.23 |  2841525 |
| pedestrian       |      67.25 | 42.22 |   248071 |
| bike             |      22.61 | 15.19 |    98087 |
| traffic_boundary |      70.63 | 50.87 |   376372 |
| background       |      98.78 | 98.27 | 85455437 |
+------------------+------------+-------+----------+
2022-10-25 19:07:09,814 adals.validate INFO: 3D overall accuracy=98.36%
2022-10-25 19:07:09,814 adals.validate INFO: 3D overall IOU=60.94
2022-10-25 19:07:09,815 adals.validate INFO: 3D class-wise segmentation accuracy and IoU.
+------------------+------------+-------+----------+
| Class            |   Accuracy |   IOU |    Total |
|------------------+------------+-------+----------|
| vehicle          |      91.33 | 70.83 |  2605498 |
| pedestrian       |      76.99 | 56.14 |   244248 |
| bike             |      58.46 | 42.18 |    92887 |
| traffic_boundary |      79.38 | 37.17 |   388136 |
| background       |      98.71 | 98.36 | 98362495 |
+------------------+------------+-------+----------+
2022-10-25 19:07:09,815 adals.validate INFO: 2D+3D overall accuracy=99.08%
2022-10-25 19:07:09,816 adals.validate INFO: 2D+3D overall IOU=66.36
2022-10-25 19:07:09,816 adals.validate INFO: 2D+3D class-wise segmentation accuracy and IoU.
+------------------+------------+-------+----------+
| Class            |   Accuracy |   IOU |    Total |
|------------------+------------+-------+----------|
| vehicle          |      92.01 | 80.58 |  2605498 |
| pedestrian       |      75.36 | 61.51 |   244248 |
| bike             |      36.67 | 32.18 |    92887 |
| traffic_boundary |      79.47 | 58.43 |   388136 |
| background       |      99.46 | 99.10 | 98362495 |
+------------------+------------+-------+----------+
