search

open-mmlab / OpenPCDet

OpenPCDet Toolbox for LiDAR-based 3D Object Detection.
Apache License 2.0
4.5k stars 1.27k forks source link

Much lower detection results on Kitti Pedestrian class + error setting USE_ROAD_PLANE = True #1580

Closed sinatayebati closed 3 months ago

sinatayebati commented 4 months ago

I have done an experiment with training SECOND model on kitti, but for some reason when I use the default config at kitti_dataset.yaml, I get the following error:

File "/home/sina/miniconda3/envs/rd-mae/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py", line 49, in data = [self.dataset[idx] for idx in possibly_batched_index] File "../pcdet/datasets/kitti/kitti_dataset.py", line 425, in getitem data_dict = self.prepare_data(data_dict=input_dict) File "../pcdet/datasets/dataset.py", line 188, in prepare_data 'gt_boxes_mask': gt_boxes_mask File "../pcdet/datasets/augmentor/data_augmentor.py", line 302, in forward data_dict = cur_augmentor(data_dict=data_dict) File "../pcdet/datasets/augmentor/database_sampler.py", line 498, in call data_dict, sampled_gt_boxes, total_valid_sampled_dict, sampled_mv_height, sampled_gt_boxes2d File "../pcdet/datasets/augmentor/database_sampler.py", line 372, in add_sampled_boxes_to_scene sampled_gt_boxes, data_dict['road_plane'], data_dict['calib']

This is despite the fact that I'm using the "planes" file for producing data infos as stated in the GETTING STARTED. After facing this error, I set the USE_ROAD_PLANE to "False" to be able to execute the training. Although the training goes through, the AP results are very different from the report in particular for "pedestrian" class.

1 - does setting USE_ROAD_PLANE to False in DATA AUGMENTATION cause this much of change in detection accuracy of pedestrian class? 2 - what could be the cause of facing the error in in the first place when using the default config and can I resolve this error?

Following is the log of my training.

2024-03-07 08:39:22,448 INFO **Start logging** 2024-03-07 08:39:22,448 INFO CUDA_VISIBLE_DEVICES=ALL 2024-03-07 08:39:22,448 INFO Training in distributed mode : total_batch_size: 8 2024-03-07 08:39:22,448 INFO cfg_file cfgs/kitti_models/second.yaml 2024-03-07 08:39:22,448 INFO batch_size 4 2024-03-07 08:39:22,448 INFO epochs 80 2024-03-07 08:39:22,448 INFO workers 4 2024-03-07 08:39:22,448 INFO extra_tag default 2024-03-07 08:39:22,448 INFO ckpt None 2024-03-07 08:39:22,448 INFO pretrained_model None 2024-03-07 08:39:22,448 INFO launcher pytorch 2024-03-07 08:39:22,448 INFO tcp_port 18888 2024-03-07 08:39:22,448 INFO sync_bn False 2024-03-07 08:39:22,448 INFO fix_random_seed False 2024-03-07 08:39:22,448 INFO ckpt_save_interval 1 2024-03-07 08:39:22,448 INFO local_rank 0 2024-03-07 08:39:22,448 INFO max_ckpt_save_num 15 2024-03-07 08:39:22,448 INFO merge_all_iters_to_one_epoch False 2024-03-07 08:39:22,448 INFO set_cfgs None 2024-03-07 08:39:22,448 INFO max_waiting_mins 0 2024-03-07 08:39:22,448 INFO start_epoch 0 2024-03-07 08:39:22,448 INFO num_epochs_to_eval 10 2024-03-07 08:39:22,448 INFO save_to_file False 2024-03-07 08:39:22,448 INFO use_tqdm_to_record False 2024-03-07 08:39:22,448 INFO logger_iter_interval 50 2024-03-07 08:39:22,449 INFO ckpt_save_time_interval 300 2024-03-07 08:39:22,449 INFO wo_gpu_stat False 2024-03-07 08:39:22,449 INFO use_amp False 2024-03-07 08:39:22,449 INFO cfg.ROOT_DIR: /hdd_10tb/sina/Radial_MAE 2024-03-07 08:39:22,449 INFO cfg.LOCAL_RANK: 0 2024-03-07 08:39:22,449 INFO cfg.CLASS_NAMES: ['Car', 'Pedestrian', 'Cyclist'] 2024-03-07 08:39:22,449 INFO ----------- DATA_CONFIG ----------- 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.DATASET: KittiDataset 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.DATA_PATH: ../data/kitti 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.POINT_CLOUD_RANGE: [0, -40, -3, 70.4, 40, 1] 2024-03-07 08:39:22,449 INFO ----------- DATA_SPLIT ----------- 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.DATA_SPLIT.train: train 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.DATA_SPLIT.test: val 2024-03-07 08:39:22,449 INFO ----------- INFO_PATH ----------- 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.INFO_PATH.train: ['kitti_infos_train.pkl'] 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.INFO_PATH.test: ['kitti_infos_val.pkl'] 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.GET_ITEM_LIST: ['points'] 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.FOV_POINTS_ONLY: True 2024-03-07 08:39:22,449 INFO ----------- DATA_AUGMENTOR ----------- 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.DATA_AUGMENTOR.DISABLE_AUG_LIST: ['placeholder'] 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.DATA_AUGMENTOR.AUG_CONFIG_LIST: [{'NAME': 'gt_sampling', 'USE_ROAD_PLANE': False, 'DB_INFO_PATH': ['kitti_dbinfos_train.pkl'], 'PREPARE': {'filter_by_min_points': ['Car:5', 'Pedestrian:5', 'Cyclist:5'], 'filter_by_difficulty': [-1]}, 'SAMPLE_GROUPS': ['Car:20', 'Pedestrian:15', 'Cyclist:15'], 'NUM_POINT_FEATURES': 4, 'DATABASE_WITH_FAKELIDAR': False, 'REMOVE_EXTRA_WIDTH': [0.0, 0.0, 0.0], 'LIMIT_WHOLE_SCENE': True}, {'NAME': 'random_world_flip', 'ALONG_AXIS_LIST': ['x']}, {'NAME': 'random_world_rotation', 'WORLD_ROT_ANGLE': [-0.78539816, 0.78539816]}, {'NAME': 'random_world_scaling', 'WORLD_SCALE_RANGE': [0.95, 1.05]}] 2024-03-07 08:39:22,449 INFO ----------- POINT_FEATURE_ENCODING ----------- 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.POINT_FEATURE_ENCODING.encoding_type: absolute_coordinates_encoding 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.POINT_FEATURE_ENCODING.used_feature_list: ['x', 'y', 'z', 'intensity'] 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.POINT_FEATURE_ENCODING.src_feature_list: ['x', 'y', 'z', 'intensity'] 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG.DATA_PROCESSOR: [{'NAME': 'mask_points_and_boxes_outside_range', 'REMOVE_OUTSIDE_BOXES': True}, {'NAME': 'shuffle_points', 'SHUFFLE_ENABLED': {'train': True, 'test': False}}, {'NAME': 'transform_points_to_voxels', 'VOXEL_SIZE': [0.05, 0.05, 0.1], 'MAX_POINTS_PER_VOXEL': 5, 'MAX_NUMBER_OF_VOXELS': {'train': 16000, 'test': 40000}}] 2024-03-07 08:39:22,449 INFO cfg.DATA_CONFIG._BASECONFIG: cfgs/dataset_configs/kitti_dataset.yaml 2024-03-07 08:39:22,449 INFO ----------- MODEL ----------- 2024-03-07 08:39:22,449 INFO cfg.MODEL.NAME: SECONDNet 2024-03-07 08:39:22,449 INFO ----------- VFE ----------- 2024-03-07 08:39:22,449 INFO cfg.MODEL.VFE.NAME: MeanVFE 2024-03-07 08:39:22,449 INFO ----------- BACKBONE_3D ----------- 2024-03-07 08:39:22,449 INFO cfg.MODEL.BACKBONE_3D.NAME: VoxelBackBone8x 2024-03-07 08:39:22,449 INFO ----------- MAP_TO_BEV ----------- 2024-03-07 08:39:22,449 INFO cfg.MODEL.MAP_TO_BEV.NAME: HeightCompression 2024-03-07 08:39:22,449 INFO cfg.MODEL.MAP_TO_BEV.NUM_BEV_FEATURES: 256 2024-03-07 08:39:22,449 INFO ----------- BACKBONE_2D ----------- 2024-03-07 08:39:22,450 INFO cfg.MODEL.BACKBONE_2D.NAME: BaseBEVBackbone 2024-03-07 08:39:22,450 INFO cfg.MODEL.BACKBONE_2D.LAYER_NUMS: [5, 5] 2024-03-07 08:39:22,450 INFO cfg.MODEL.BACKBONE_2D.LAYER_STRIDES: [1, 2] 2024-03-07 08:39:22,450 INFO cfg.MODEL.BACKBONE_2D.NUM_FILTERS: [128, 256] 2024-03-07 08:39:22,450 INFO cfg.MODEL.BACKBONE_2D.UPSAMPLE_STRIDES: [1, 2] 2024-03-07 08:39:22,450 INFO cfg.MODEL.BACKBONE_2D.NUM_UPSAMPLE_FILTERS: [256, 256] 2024-03-07 08:39:22,450 INFO ----------- DENSE_HEAD ----------- 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.NAME: AnchorHeadSingle 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.CLASS_AGNOSTIC: False 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.USE_DIRECTION_CLASSIFIER: True 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.DIR_OFFSET: 0.78539 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.DIR_LIMIT_OFFSET: 0.0 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.NUM_DIR_BINS: 2 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.ANCHOR_GENERATOR_CONFIG: [{'class_name': 'Car', 'anchor_sizes': [[3.9, 1.6, 1.56]], 'anchor_rotations': [0, 1.57], 'anchor_bottom_heights': [-1.78], 'align_center': False, 'feature_map_stride': 8, 'matched_threshold': 0.6, 'unmatched_threshold': 0.45}, {'class_name': 'Pedestrian', 'anchor_sizes': [[0.8, 0.6, 1.73]], 'anchor_rotations': [0, 1.57], 'anchor_bottom_heights': [-0.6], 'align_center': False, 'feature_map_stride': 8, 'matched_threshold': 0.5, 'unmatched_threshold': 0.35}, {'class_name': 'Cyclist', 'anchor_sizes': [[1.76, 0.6, 1.73]], 'anchor_rotations': [0, 1.57], 'anchor_bottom_heights': [-0.6], 'align_center': False, 'feature_map_stride': 8, 'matched_threshold': 0.5, 'unmatched_threshold': 0.35}] 2024-03-07 08:39:22,450 INFO ----------- TARGET_ASSIGNER_CONFIG ----------- 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.TARGET_ASSIGNER_CONFIG.NAME: AxisAlignedTargetAssigner 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.TARGET_ASSIGNER_CONFIG.POS_FRACTION: -1.0 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.TARGET_ASSIGNER_CONFIG.SAMPLE_SIZE: 512 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.TARGET_ASSIGNER_CONFIG.NORM_BY_NUM_EXAMPLES: False 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.TARGET_ASSIGNER_CONFIG.MATCH_HEIGHT: False 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.TARGET_ASSIGNER_CONFIG.BOX_CODER: ResidualCoder 2024-03-07 08:39:22,450 INFO ----------- LOSS_CONFIG ----------- 2024-03-07 08:39:22,450 INFO ----------- LOSS_WEIGHTS ----------- 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.LOSS_CONFIG.LOSS_WEIGHTS.cls_weight: 1.0 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.LOSS_CONFIG.LOSS_WEIGHTS.loc_weight: 2.0 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.LOSS_CONFIG.LOSS_WEIGHTS.dir_weight: 0.2 2024-03-07 08:39:22,450 INFO cfg.MODEL.DENSE_HEAD.LOSS_CONFIG.LOSS_WEIGHTS.code_weights: [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0] 2024-03-07 08:39:22,450 INFO ----------- POST_PROCESSING ----------- 2024-03-07 08:39:22,450 INFO cfg.MODEL.POST_PROCESSING.RECALL_THRESH_LIST: [0.3, 0.5, 0.7] 2024-03-07 08:39:22,450 INFO cfg.MODEL.POST_PROCESSING.SCORE_THRESH: 0.1 2024-03-07 08:39:22,450 INFO cfg.MODEL.POST_PROCESSING.OUTPUT_RAW_SCORE: False 2024-03-07 08:39:22,450 INFO cfg.MODEL.POST_PROCESSING.EVAL_METRIC: kitti 2024-03-07 08:39:22,450 INFO ----------- NMS_CONFIG ----------- 2024-03-07 08:39:22,450 INFO cfg.MODEL.POST_PROCESSING.NMS_CONFIG.MULTI_CLASSES_NMS: False 2024-03-07 08:39:22,450 INFO cfg.MODEL.POST_PROCESSING.NMS_CONFIG.NMS_TYPE: nms_gpu 2024-03-07 08:39:22,450 INFO cfg.MODEL.POST_PROCESSING.NMS_CONFIG.NMS_THRESH: 0.01 2024-03-07 08:39:22,450 INFO cfg.MODEL.POST_PROCESSING.NMS_CONFIG.NMS_PRE_MAXSIZE: 4096 2024-03-07 08:39:22,450 INFO cfg.MODEL.POST_PROCESSING.NMS_CONFIG.NMS_POST_MAXSIZE: 500 2024-03-07 08:39:22,451 INFO ----------- OPTIMIZATION ----------- 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.BATCH_SIZE_PER_GPU: 4 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.NUM_EPOCHS: 80 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.OPTIMIZER: adam_onecycle 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.LR: 0.003 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.WEIGHT_DECAY: 0.01 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.MOMENTUM: 0.9 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.MOMS: [0.95, 0.85] 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.PCT_START: 0.4 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.DIV_FACTOR: 10 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.DECAY_STEP_LIST: [35, 45] 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.LR_DECAY: 0.1 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.LR_CLIP: 1e-07 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.LR_WARMUP: False 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.WARMUP_EPOCH: 1 2024-03-07 08:39:22,451 INFO cfg.OPTIMIZATION.GRAD_NORM_CLIP: 10 2024-03-07 08:39:22,451 INFO cfg.TAG: second 2024-03-07 08:39:22,451 INFO cfg.EXP_GROUP_PATH: kitti_models 2024-03-07 08:39:22,464 INFO ----------- Create dataloader & network & optimizer ----------- 2024-03-07 08:39:22,574 INFO Database filter by min points Car: 14357 => 13532 2024-03-07 08:39:22,574 INFO Database filter by min points Pedestrian: 2207 => 2168 2024-03-07 08:39:22,574 INFO Database filter by min points Cyclist: 734 => 705 2024-03-07 08:39:22,591 INFO Database filter by difficulty Car: 13532 => 10759 2024-03-07 08:39:22,594 INFO Database filter by difficulty Pedestrian: 2168 => 2075 2024-03-07 08:39:22,595 INFO Database filter by difficulty Cyclist: 705 => 581 2024-03-07 08:39:22,601 INFO Loading KITTI dataset 2024-03-07 08:39:22,689 INFO Total samples for KITTI dataset: 3712 2024-03-07 08:39:35,808 INFO ----------- Model SECONDNet created, param count: 5325576 ----------- 2024-03-07 08:39:35,808 INFO DistributedDataParallel( (module): SECONDNet( (vfe): MeanVFE() (backbone_3d): VoxelBackBone8x( (conv_input): SparseSequential( (0): SubMConv3d(4, 16, kernel_size=[3, 3, 3], stride=[1, 1, 1], padding=[1, 1, 1], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(16, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) (conv1): SparseSequential( (0): SparseSequential( (0): SubMConv3d(16, 16, kernel_size=[3, 3, 3], stride=[1, 1, 1], padding=[0, 0, 0], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(16, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) ) (conv2): SparseSequential( (0): SparseSequential( (0): SparseConv3d(16, 32, kernel_size=[3, 3, 3], stride=[2, 2, 2], padding=[1, 1, 1], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(32, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) (1): SparseSequential( (0): SubMConv3d(32, 32, kernel_size=[3, 3, 3], stride=[1, 1, 1], padding=[0, 0, 0], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(32, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) (2): SparseSequential( (0): SubMConv3d(32, 32, kernel_size=[3, 3, 3], stride=[1, 1, 1], padding=[0, 0, 0], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(32, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) ) (conv3): SparseSequential( (0): SparseSequential( (0): SparseConv3d(32, 64, kernel_size=[3, 3, 3], stride=[2, 2, 2], padding=[1, 1, 1], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) (1): SparseSequential( (0): SubMConv3d(64, 64, kernel_size=[3, 3, 3], stride=[1, 1, 1], padding=[0, 0, 0], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) (2): SparseSequential( (0): SubMConv3d(64, 64, kernel_size=[3, 3, 3], stride=[1, 1, 1], padding=[0, 0, 0], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) ) (conv4): SparseSequential( (0): SparseSequential( (0): SparseConv3d(64, 64, kernel_size=[3, 3, 3], stride=[2, 2, 2], padding=[0, 1, 1], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) (1): SparseSequential( (0): SubMConv3d(64, 64, kernel_size=[3, 3, 3], stride=[1, 1, 1], padding=[0, 0, 0], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) (2): SparseSequential( (0): SubMConv3d(64, 64, kernel_size=[3, 3, 3], stride=[1, 1, 1], padding=[0, 0, 0], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(64, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) ) (conv_out): SparseSequential( (0): SparseConv3d(64, 128, kernel_size=[3, 1, 1], stride=[2, 1, 1], padding=[0, 0, 0], dilation=[1, 1, 1], output_padding=[0, 0, 0], bias=False, algo=ConvAlgo.MaskImplicitGemm) (1): BatchNorm1d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) ) (map_to_bev_module): HeightCompression() (pfe): None (backbone_2d): BaseBEVBackbone( (blocks): ModuleList( (0): Sequential( (0): ZeroPad2d(padding=(1, 1, 1, 1), value=0.0) (1): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), bias=False) (2): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (3): ReLU() (4): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (5): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (6): ReLU() (7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (8): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (9): ReLU() (10): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (11): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (12): ReLU() (13): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (14): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (15): ReLU() (16): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (17): BatchNorm2d(128, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (18): ReLU() ) (1): Sequential( (0): ZeroPad2d(padding=(1, 1, 1, 1), value=0.0) (1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), bias=False) (2): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (3): ReLU() (4): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (5): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (6): ReLU() (7): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (8): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (9): ReLU() (10): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (11): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (12): ReLU() (13): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (14): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (15): ReLU() (16): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (17): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (18): ReLU() ) ) (deblocks): ModuleList( (0): Sequential( (0): ConvTranspose2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False) (1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) (1): Sequential( (0): ConvTranspose2d(256, 256, kernel_size=(2, 2), stride=(2, 2), bias=False) (1): BatchNorm2d(256, eps=0.001, momentum=0.01, affine=True, track_running_stats=True) (2): ReLU() ) ) ) (dense_head): AnchorHeadSingle( (cls_loss_func): SigmoidFocalClassificationLoss() (reg_loss_func): WeightedSmoothL1Loss() (dir_loss_func): WeightedCrossEntropyLoss() (conv_cls): Conv2d(512, 18, kernel_size=(1, 1), stride=(1, 1)) (conv_box): Conv2d(512, 42, kernel_size=(1, 1), stride=(1, 1)) (conv_dir_cls): Conv2d(512, 12, kernel_size=(1, 1), stride=(1, 1)) ) (point_head): None (roi_head): None ) ) 2024-03-07 08:39:35,810 INFO **Start training kitti_models/second(default)** 2024-03-07 13:28:01,976 INFO **End training kitti_models/second(default)**

2024-03-07 13:28:01,976 INFO **Start evaluation kitti_models/second(default)** 2024-03-07 13:28:01,978 INFO Loading KITTI dataset 2024-03-07 13:28:02,095 INFO Total samples for KITTI dataset: 3769 2024-03-07 13:28:02,097 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_70.pth to CPU 2024-03-07 13:28:02,129 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:28:02,156 INFO ==> Done (loaded 163/163) 2024-03-07 13:28:02,162 INFO * EPOCH 70 EVALUATION *** 2024-03-07 13:29:16,876 INFO * Performance of EPOCH 70 *** 2024-03-07 13:29:16,876 INFO Generate label finished(sec_per_example: 0.0198 second). 2024-03-07 13:29:16,876 INFO recall_roi_0.3: 0.000000 2024-03-07 13:29:16,876 INFO recall_rcnn_0.3: 0.935820 2024-03-07 13:29:16,876 INFO recall_roi_0.5: 0.000000 2024-03-07 13:29:16,876 INFO recall_rcnn_0.5: 0.873747 2024-03-07 13:29:16,876 INFO recall_roi_0.7: 0.000000 2024-03-07 13:29:16,876 INFO recall_rcnn_0.7: 0.641800 2024-03-07 13:29:16,879 INFO Average predicted number of objects(3769 samples): 15.882 2024-03-07 13:29:41,927 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7287, 89.3570, 88.1035 bev AP:89.7277, 86.5316, 83.7969 3d AP:87.2732, 77.2367, 75.0465 aos AP:90.71, 89.17, 87.78 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.4995, 91.7337, 90.4806 bev AP:92.0201, 87.5550, 86.0731 3d AP:89.1748, 78.4544, 75.2503 aos AP:95.48, 91.52, 90.12 Car AP@0.70, 0.50, 0.50: bbox AP:90.7287, 89.3570, 88.1035 bev AP:90.7364, 89.8785, 88.9184 3d AP:90.7331, 89.8099, 88.7457 aos AP:90.71, 89.17, 87.78 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.4995, 91.7337, 90.4806 bev AP:95.5009, 94.2510, 93.2576 3d AP:95.4804, 93.9832, 92.5836 aos AP:95.48, 91.52, 90.12 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:67.3698, 63.1083, 59.5613 bev AP:55.5809, 51.2184, 46.6118 3d AP:51.5930, 46.1262, 42.0066 aos AP:63.34, 58.19, 54.59 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:67.5259, 62.8909, 58.9326 bev AP:55.3937, 49.9824, 45.4560 3d AP:50.6909, 45.0838, 40.2616 aos AP:62.99, 57.37, 53.38 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:67.3698, 63.1083, 59.5613 bev AP:72.6992, 69.1130, 65.8344 3d AP:72.3941, 68.4985, 64.5503 aos AP:63.34, 58.19, 54.59 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:67.5259, 62.8909, 58.9326 bev AP:73.0713, 69.7680, 65.8169 3d AP:72.8027, 69.3177, 65.1054 aos AP:62.99, 57.37, 53.38 Cyclist AP@0.50, 0.50, 0.50: bbox AP:85.1467, 76.4681, 72.5613 bev AP:82.0800, 69.5228, 64.9106 3d AP:79.2070, 66.5490, 61.8594 aos AP:84.76, 75.57, 71.72 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:88.6735, 78.0640, 73.6426 bev AP:84.6564, 70.2510, 65.5573 3d AP:80.4341, 65.9226, 61.6909 aos AP:88.22, 77.10, 72.74 Cyclist AP@0.50, 0.25, 0.25: bbox AP:85.1467, 76.4681, 72.5613 bev AP:86.2685, 74.5024, 70.2665 3d AP:86.2685, 74.5024, 70.2665 aos AP:84.76, 75.57, 71.72 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:88.6735, 78.0640, 73.6426 bev AP:88.7139, 75.5838, 71.2750 3d AP:88.7139, 75.5838, 71.2750 aos AP:88.22, 77.10, 72.74

2024-03-07 13:29:41,934 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_70/val 2024-03-07 13:29:41,934 INFO ****Evaluation done.** 2024-03-07 13:29:41,953 INFO Epoch 70 has been evaluated 2024-03-07 13:29:41,954 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_71.pth to CPU 2024-03-07 13:29:41,980 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:29:42,293 INFO ==> Done (loaded 163/163) 2024-03-07 13:29:42,295 INFO EPOCH 71 EVALUATION ** 2024-03-07 13:30:56,656 INFO Performance of EPOCH 71 ***** 2024-03-07 13:30:56,656 INFO Generate label finished(sec_per_example: 0.0197 second). 2024-03-07 13:30:56,656 INFO recall_roi_0.3: 0.000000 2024-03-07 13:30:56,656 INFO recall_rcnn_0.3: 0.935820 2024-03-07 13:30:56,656 INFO recall_roi_0.5: 0.000000 2024-03-07 13:30:56,657 INFO recall_rcnn_0.5: 0.870615 2024-03-07 13:30:56,657 INFO recall_roi_0.7: 0.000000 2024-03-07 13:30:56,657 INFO recall_rcnn_0.7: 0.636788 2024-03-07 13:30:56,659 INFO Average predicted number of objects(3769 samples): 15.607 2024-03-07 13:31:15,014 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7828, 89.5906, 88.4399 bev AP:89.7004, 86.3921, 84.4322 3d AP:87.6256, 77.3643, 75.3129 aos AP:90.77, 89.42, 88.13 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.6302, 91.9791, 90.9312 bev AP:91.9936, 87.4932, 86.3728 3d AP:88.3678, 78.5168, 75.4792 aos AP:95.61, 91.78, 90.58 Car AP@0.70, 0.50, 0.50: bbox AP:90.7828, 89.5906, 88.4399 bev AP:90.7867, 89.9605, 89.1172 3d AP:90.7867, 89.8827, 88.9732 aos AP:90.77, 89.42, 88.13 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.6302, 91.9791, 90.9312 bev AP:95.6565, 94.4343, 93.5962 3d AP:95.6398, 94.1747, 93.0450 aos AP:95.61, 91.78, 90.58 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:68.6141, 64.4765, 60.9285 bev AP:56.6676, 52.7499, 48.4110 3d AP:51.0755, 46.0687, 41.2724 aos AP:64.75, 59.90, 56.35 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:68.9825, 64.4930, 60.7010 bev AP:56.5868, 51.2196, 46.5553 3d AP:49.6928, 44.2345, 39.4647 aos AP:64.76, 59.59, 55.71 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:68.6141, 64.4765, 60.9285 bev AP:74.0876, 70.6416, 66.8259 3d AP:73.8675, 70.3349, 66.6118 aos AP:64.75, 59.90, 56.35 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:68.9825, 64.4930, 60.7010 bev AP:74.9806, 71.2599, 67.1094 3d AP:74.6523, 70.8983, 66.6582 aos AP:64.76, 59.59, 55.71 Cyclist AP@0.50, 0.50, 0.50: bbox AP:86.6304, 78.6902, 75.1744 bev AP:83.0774, 71.9258, 67.1078 3d AP:81.9831, 68.6729, 63.8713 aos AP:86.45, 77.56, 74.06 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:90.6653, 80.0295, 76.0954 bev AP:86.4530, 72.4607, 67.3143 3d AP:85.0338, 68.9957, 64.0409 aos AP:90.43, 78.77, 74.87 Cyclist AP@0.50, 0.25, 0.25: bbox AP:86.6304, 78.6902, 75.1744 bev AP:89.3790, 76.6739, 73.1732 3d AP:89.3790, 76.6739, 73.1732 aos AP:86.45, 77.56, 74.06 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:90.6653, 80.0295, 76.0954 bev AP:91.3734, 77.7785, 74.0852 3d AP:91.3734, 77.7785, 74.0852 aos AP:90.43, 78.77, 74.87

2024-03-07 13:31:15,018 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_71/val 2024-03-07 13:31:15,018 INFO ****Evaluation done.** 2024-03-07 13:31:15,037 INFO Epoch 71 has been evaluated 2024-03-07 13:31:15,038 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_72.pth to CPU 2024-03-07 13:31:15,064 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:31:15,376 INFO ==> Done (loaded 163/163) 2024-03-07 13:31:15,378 INFO EPOCH 72 EVALUATION ** 2024-03-07 13:32:33,366 INFO Performance of EPOCH 72 ***** 2024-03-07 13:32:33,367 INFO Generate label finished(sec_per_example: 0.0207 second). 2024-03-07 13:32:33,367 INFO recall_roi_0.3: 0.000000 2024-03-07 13:32:33,367 INFO recall_rcnn_0.3: 0.938440 2024-03-07 13:32:33,367 INFO recall_roi_0.5: 0.000000 2024-03-07 13:32:33,367 INFO recall_rcnn_0.5: 0.874943 2024-03-07 13:32:33,367 INFO recall_roi_0.7: 0.000000 2024-03-07 13:32:33,367 INFO recall_rcnn_0.7: 0.642084 2024-03-07 13:32:33,369 INFO Average predicted number of objects(3769 samples): 16.015 2024-03-07 13:32:49,954 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7770, 89.5740, 88.4705 bev AP:89.7529, 86.7814, 84.9862 3d AP:87.4309, 77.3947, 75.4585 aos AP:90.75, 89.39, 88.16 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.4914, 91.9809, 90.9509 bev AP:91.9535, 87.6785, 86.6012 3d AP:88.1028, 78.5590, 75.5222 aos AP:95.46, 91.78, 90.62 Car AP@0.70, 0.50, 0.50: bbox AP:90.7770, 89.5740, 88.4705 bev AP:90.7659, 89.9374, 89.1149 3d AP:90.7659, 89.8541, 88.9458 aos AP:90.75, 89.39, 88.16 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.4914, 91.9809, 90.9509 bev AP:95.5182, 94.4416, 93.6070 3d AP:95.5042, 94.2000, 93.1159 aos AP:95.46, 91.78, 90.62 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:68.1828, 64.1627, 60.1360 bev AP:55.8984, 52.3943, 48.1191 3d AP:51.6460, 46.4825, 42.4019 aos AP:62.71, 58.31, 54.18 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:68.4762, 64.1739, 59.9450 bev AP:55.7714, 50.7975, 46.0914 3d AP:50.0573, 44.7613, 40.1573 aos AP:62.57, 57.69, 53.45 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:68.1828, 64.1627, 60.1360 bev AP:74.3217, 70.7955, 66.7539 3d AP:74.0430, 70.4357, 66.4592 aos AP:62.71, 58.31, 54.18 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:68.4762, 64.1739, 59.9450 bev AP:75.1699, 71.6200, 67.1233 3d AP:74.8101, 71.2009, 66.6752 aos AP:62.57, 57.69, 53.45 Cyclist AP@0.50, 0.50, 0.50: bbox AP:87.0778, 78.7818, 74.0592 bev AP:83.3260, 71.6391, 67.7179 3d AP:82.1782, 69.6177, 64.7907 aos AP:86.79, 78.05, 73.29 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:91.5899, 80.2099, 75.5422 bev AP:87.3028, 72.6901, 67.9401 3d AP:84.5839, 69.8799, 65.0377 aos AP:91.25, 79.36, 74.73 Cyclist AP@0.50, 0.25, 0.25: bbox AP:87.0778, 78.7818, 74.0592 bev AP:90.9604, 76.7338, 72.0969 3d AP:90.9512, 76.7232, 72.0790 aos AP:86.79, 78.05, 73.29 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:91.5899, 80.2099, 75.5422 bev AP:92.0846, 77.9207, 73.6115 3d AP:92.0699, 77.9066, 73.5974 aos AP:91.25, 79.36, 74.73

2024-03-07 13:32:49,958 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_72/val 2024-03-07 13:32:49,958 INFO ****Evaluation done.** 2024-03-07 13:32:49,978 INFO Epoch 72 has been evaluated 2024-03-07 13:32:49,979 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_73.pth to CPU 2024-03-07 13:32:50,006 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:32:50,320 INFO ==> Done (loaded 163/163) 2024-03-07 13:32:50,322 INFO EPOCH 73 EVALUATION ** 2024-03-07 13:34:07,056 INFO Performance of EPOCH 73 ***** 2024-03-07 13:34:07,056 INFO Generate label finished(sec_per_example: 0.0204 second). 2024-03-07 13:34:07,056 INFO recall_roi_0.3: 0.000000 2024-03-07 13:34:07,056 INFO recall_rcnn_0.3: 0.936845 2024-03-07 13:34:07,056 INFO recall_roi_0.5: 0.000000 2024-03-07 13:34:07,056 INFO recall_rcnn_0.5: 0.875740 2024-03-07 13:34:07,056 INFO recall_roi_0.7: 0.000000 2024-03-07 13:34:07,056 INFO recall_rcnn_0.7: 0.641856 2024-03-07 13:34:07,058 INFO Average predicted number of objects(3769 samples): 15.748 2024-03-07 13:34:23,533 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7721, 89.5417, 88.3915 bev AP:89.7840, 86.6944, 84.6755 3d AP:87.3223, 77.3723, 75.3097 aos AP:90.76, 89.35, 88.07 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.6713, 92.0924, 90.9339 bev AP:92.1975, 87.7529, 86.5425 3d AP:88.2052, 78.6150, 75.5605 aos AP:95.65, 91.89, 90.58 Car AP@0.70, 0.50, 0.50: bbox AP:90.7721, 89.5417, 88.3915 bev AP:90.7643, 89.9042, 89.0915 3d AP:90.7643, 89.8365, 88.9442 aos AP:90.76, 89.35, 88.07 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.6713, 92.0924, 90.9339 bev AP:95.6973, 94.4750, 93.5806 3d AP:95.6802, 94.2453, 93.0040 aos AP:95.65, 91.89, 90.58 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:68.6540, 64.2126, 60.2155 bev AP:57.2683, 53.2045, 48.5469 3d AP:51.3736, 47.1749, 42.5936 aos AP:64.65, 59.97, 55.72 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:68.6487, 63.9813, 60.0315 bev AP:57.8007, 52.3044, 47.4778 3d AP:50.4774, 45.8073, 40.8765 aos AP:64.27, 59.25, 54.96 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:68.6540, 64.2126, 60.2155 bev AP:74.2580, 70.2737, 66.7267 3d AP:74.0631, 69.9129, 66.5078 aos AP:64.65, 59.97, 55.72 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:68.6487, 63.9813, 60.0315 bev AP:75.0382, 70.8805, 66.7598 3d AP:74.7014, 70.5595, 66.4486 aos AP:64.27, 59.25, 54.96 Cyclist AP@0.50, 0.50, 0.50: bbox AP:86.6842, 78.7651, 74.5724 bev AP:83.6757, 71.9338, 67.7737 3d AP:81.7559, 68.5597, 63.9543 aos AP:86.43, 77.92, 73.75 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:90.7675, 80.1333, 75.8307 bev AP:87.1972, 72.9084, 68.2191 3d AP:84.7066, 68.8114, 64.3369 aos AP:90.45, 79.18, 74.91 Cyclist AP@0.50, 0.25, 0.25: bbox AP:86.6842, 78.7651, 74.5724 bev AP:89.9852, 76.8165, 72.5484 3d AP:89.9852, 76.8165, 72.5484 aos AP:86.43, 77.92, 73.75 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:90.7675, 80.1333, 75.8307 bev AP:91.1968, 77.8988, 73.5349 3d AP:91.1968, 77.8988, 73.5349 aos AP:90.45, 79.18, 74.91

2024-03-07 13:34:23,537 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_73/val 2024-03-07 13:34:23,537 INFO ****Evaluation done.** 2024-03-07 13:34:23,555 INFO Epoch 73 has been evaluated 2024-03-07 13:34:23,556 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_74.pth to CPU 2024-03-07 13:34:23,583 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:34:23,896 INFO ==> Done (loaded 163/163) 2024-03-07 13:34:23,898 INFO EPOCH 74 EVALUATION ** 2024-03-07 13:35:37,978 INFO Performance of EPOCH 74 ***** 2024-03-07 13:35:37,978 INFO Generate label finished(sec_per_example: 0.0197 second). 2024-03-07 13:35:37,978 INFO recall_roi_0.3: 0.000000 2024-03-07 13:35:37,978 INFO recall_rcnn_0.3: 0.936503 2024-03-07 13:35:37,978 INFO recall_roi_0.5: 0.000000 2024-03-07 13:35:37,978 INFO recall_rcnn_0.5: 0.873747 2024-03-07 13:35:37,978 INFO recall_roi_0.7: 0.000000 2024-03-07 13:35:37,978 INFO recall_rcnn_0.7: 0.639636 2024-03-07 13:35:37,980 INFO Average predicted number of objects(3769 samples): 15.682 2024-03-07 13:35:54,412 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7833, 89.5242, 88.3915 bev AP:89.7806, 86.5216, 83.9099 3d AP:87.5900, 77.4168, 75.4179 aos AP:90.76, 89.35, 88.09 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.5559, 91.9137, 90.8014 bev AP:92.0375, 87.5690, 86.2494 3d AP:88.3245, 78.6058, 75.5374 aos AP:95.53, 91.72, 90.47 Car AP@0.70, 0.50, 0.50: bbox AP:90.7833, 89.5242, 88.3915 bev AP:90.7799, 89.9356, 89.0766 3d AP:90.7799, 89.8709, 88.9225 aos AP:90.76, 89.35, 88.09 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.5559, 91.9137, 90.8014 bev AP:95.5677, 94.3951, 93.5491 3d AP:95.5525, 94.1062, 92.8987 aos AP:95.53, 91.72, 90.47 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:69.2484, 65.0338, 61.1710 bev AP:56.2730, 52.1073, 48.0030 3d AP:51.0706, 45.6070, 41.5492 aos AP:64.53, 60.05, 55.79 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:69.9680, 65.2024, 60.9936 bev AP:56.5821, 51.1293, 46.4619 3d AP:49.9444, 44.6771, 40.0048 aos AP:65.26, 60.24, 55.75 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:69.2484, 65.0338, 61.1710 bev AP:74.9060, 70.8770, 67.0464 3d AP:74.8964, 70.7381, 67.0111 aos AP:64.53, 60.05, 55.79 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:69.9680, 65.2024, 60.9936 bev AP:75.5622, 71.8084, 67.3006 3d AP:75.5497, 71.6950, 67.1532 aos AP:65.26, 60.24, 55.75 Cyclist AP@0.50, 0.50, 0.50: bbox AP:86.0468, 78.4563, 73.6232 bev AP:82.1985, 71.5314, 66.9309 3d AP:81.2401, 68.2961, 63.7010 aos AP:85.82, 77.61, 72.87 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:90.3806, 79.6812, 75.2394 bev AP:84.3407, 71.6951, 66.8858 3d AP:83.3061, 69.0294, 64.4237 aos AP:90.06, 78.79, 74.38 Cyclist AP@0.50, 0.25, 0.25: bbox AP:86.0468, 78.4563, 73.6232 bev AP:85.1634, 76.3951, 72.5143 3d AP:85.1634, 76.3951, 72.5143 aos AP:85.82, 77.61, 72.87 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:90.3806, 79.6812, 75.2394 bev AP:89.2826, 77.0069, 72.9739 3d AP:89.2826, 77.0069, 72.9739 aos AP:90.06, 78.79, 74.38

2024-03-07 13:35:54,414 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_74/val 2024-03-07 13:35:54,414 INFO ****Evaluation done.** 2024-03-07 13:35:54,434 INFO Epoch 74 has been evaluated 2024-03-07 13:35:54,435 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_75.pth to CPU 2024-03-07 13:35:54,462 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:35:54,774 INFO ==> Done (loaded 163/163) 2024-03-07 13:35:54,777 INFO EPOCH 75 EVALUATION ** 2024-03-07 13:37:09,599 INFO Performance of EPOCH 75 ***** 2024-03-07 13:37:09,600 INFO Generate label finished(sec_per_example: 0.0198 second). 2024-03-07 13:37:09,600 INFO recall_roi_0.3: 0.000000 2024-03-07 13:37:09,600 INFO recall_rcnn_0.3: 0.936959 2024-03-07 13:37:09,600 INFO recall_roi_0.5: 0.000000 2024-03-07 13:37:09,600 INFO recall_rcnn_0.5: 0.875569 2024-03-07 13:37:09,600 INFO recall_roi_0.7: 0.000000 2024-03-07 13:37:09,600 INFO recall_rcnn_0.7: 0.640831 2024-03-07 13:37:09,602 INFO Average predicted number of objects(3769 samples): 15.767 2024-03-07 13:37:26,108 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7969, 89.5382, 88.3146 bev AP:89.7544, 86.5573, 84.4824 3d AP:87.5068, 77.3232, 75.3951 aos AP:90.78, 89.36, 88.01 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.6014, 91.9435, 90.7843 bev AP:92.0420, 87.6034, 86.3541 3d AP:88.2838, 78.5696, 75.4611 aos AP:95.58, 91.75, 90.44 Car AP@0.70, 0.50, 0.50: bbox AP:90.7969, 89.5382, 88.3146 bev AP:90.7826, 89.9233, 89.0792 3d AP:90.7826, 89.8713, 88.8775 aos AP:90.78, 89.36, 88.01 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.6014, 91.9435, 90.7843 bev AP:95.6133, 94.4349, 93.5255 3d AP:95.5971, 94.2219, 92.9517 aos AP:95.58, 91.75, 90.44 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:68.7799, 64.2818, 60.4974 bev AP:58.7820, 54.3578, 49.3932 3d AP:53.3986, 48.3440, 43.8732 aos AP:64.58, 59.79, 55.85 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:69.0061, 64.3092, 60.3919 bev AP:58.4996, 52.7981, 47.9991 3d AP:52.2005, 46.7135, 41.6009 aos AP:64.48, 59.52, 55.35 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:68.7799, 64.2818, 60.4974 bev AP:74.7857, 70.8042, 67.2089 3d AP:74.5489, 70.5276, 66.9699 aos AP:64.58, 59.79, 55.85 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:69.0061, 64.3092, 60.3919 bev AP:75.4337, 71.5272, 67.3261 3d AP:75.1946, 71.2482, 66.9973 aos AP:64.48, 59.52, 55.35 Cyclist AP@0.50, 0.50, 0.50: bbox AP:86.7648, 78.6076, 74.9791 bev AP:82.6313, 71.3007, 66.9517 3d AP:81.7838, 68.5838, 63.8504 aos AP:86.50, 77.86, 74.20 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:90.9884, 79.7656, 75.6234 bev AP:84.7863, 71.9857, 66.7962 3d AP:83.8437, 69.2116, 64.5219 aos AP:90.62, 78.91, 74.79 Cyclist AP@0.50, 0.25, 0.25: bbox AP:86.7648, 78.6076, 74.9791 bev AP:90.5109, 76.2733, 72.5111 3d AP:90.5013, 76.2712, 72.5050 aos AP:86.50, 77.86, 74.20 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:90.9884, 79.7656, 75.6234 bev AP:91.1602, 77.2550, 73.1714 3d AP:91.1549, 77.2516, 73.1678 aos AP:90.62, 78.91, 74.79

2024-03-07 13:37:26,113 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_75/val 2024-03-07 13:37:26,113 INFO ****Evaluation done.** 2024-03-07 13:37:26,135 INFO Epoch 75 has been evaluated 2024-03-07 13:37:26,136 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_76.pth to CPU 2024-03-07 13:37:26,165 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:37:26,300 INFO ==> Done (loaded 163/163) 2024-03-07 13:37:26,302 INFO EPOCH 76 EVALUATION ** 2024-03-07 13:38:41,083 INFO Performance of EPOCH 76 ***** 2024-03-07 13:38:41,083 INFO Generate label finished(sec_per_example: 0.0198 second). 2024-03-07 13:38:41,083 INFO recall_roi_0.3: 0.000000 2024-03-07 13:38:41,083 INFO recall_rcnn_0.3: 0.937016 2024-03-07 13:38:41,083 INFO recall_roi_0.5: 0.000000 2024-03-07 13:38:41,083 INFO recall_rcnn_0.5: 0.875683 2024-03-07 13:38:41,083 INFO recall_roi_0.7: 0.000000 2024-03-07 13:38:41,083 INFO recall_rcnn_0.7: 0.640205 2024-03-07 13:38:41,085 INFO Average predicted number of objects(3769 samples): 16.050 2024-03-07 13:38:57,826 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7894, 89.4879, 88.3364 bev AP:89.7443, 86.6313, 84.1448 3d AP:87.5817, 77.4027, 75.4582 aos AP:90.78, 89.31, 88.01 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.5746, 91.9448, 90.7993 bev AP:92.0075, 87.6399, 86.3245 3d AP:88.2936, 78.6355, 75.5330 aos AP:95.56, 91.74, 90.44 Car AP@0.70, 0.50, 0.50: bbox AP:90.7894, 89.4879, 88.3364 bev AP:90.7819, 89.9372, 89.0757 3d AP:90.7819, 89.8793, 88.8854 aos AP:90.78, 89.31, 88.01 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.5746, 91.9448, 90.7993 bev AP:95.5890, 94.4316, 93.5349 3d AP:95.5743, 94.2002, 92.9721 aos AP:95.56, 91.74, 90.44 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:69.0194, 64.8424, 61.0892 bev AP:58.2421, 53.7427, 48.9235 3d AP:52.6994, 47.7184, 43.5128 aos AP:64.89, 60.30, 56.30 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:69.4596, 65.0059, 60.7503 bev AP:58.0152, 52.3290, 47.5070 3d AP:51.5994, 46.1229, 41.1125 aos AP:64.96, 60.04, 55.58 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:69.0194, 64.8424, 61.0892 bev AP:74.6210, 70.7917, 66.6818 3d AP:74.3146, 70.4454, 66.4469 aos AP:64.89, 60.30, 56.30 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:69.4596, 65.0059, 60.7503 bev AP:75.1698, 71.5569, 67.1665 3d AP:74.7993, 71.1523, 66.6970 aos AP:64.96, 60.04, 55.58 Cyclist AP@0.50, 0.50, 0.50: bbox AP:87.0630, 79.0385, 74.2430 bev AP:83.0270, 71.5848, 67.6187 3d AP:82.1766, 69.3762, 64.7815 aos AP:86.78, 78.28, 73.47 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:91.3093, 80.1925, 75.5488 bev AP:85.2443, 72.3984, 67.3338 3d AP:84.2224, 70.1664, 65.2989 aos AP:90.96, 79.34, 74.72 Cyclist AP@0.50, 0.25, 0.25: bbox AP:87.0630, 79.0385, 74.2430 bev AP:90.7652, 76.5908, 73.0398 3d AP:90.7652, 76.5908, 73.0398 aos AP:86.78, 78.28, 73.47 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:91.3093, 80.1925, 75.5488 bev AP:91.5675, 77.6253, 73.5250 3d AP:91.5675, 77.6253, 73.5250 aos AP:90.96, 79.34, 74.72

2024-03-07 13:38:57,827 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_76/val 2024-03-07 13:38:57,827 INFO ****Evaluation done.** 2024-03-07 13:38:57,847 INFO Epoch 76 has been evaluated 2024-03-07 13:38:57,848 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_77.pth to CPU 2024-03-07 13:38:57,875 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:38:58,188 INFO ==> Done (loaded 163/163) 2024-03-07 13:38:58,190 INFO EPOCH 77 EVALUATION ** 2024-03-07 13:40:13,746 INFO Performance of EPOCH 77 ***** 2024-03-07 13:40:13,747 INFO Generate label finished(sec_per_example: 0.0200 second). 2024-03-07 13:40:13,747 INFO recall_roi_0.3: 0.000000 2024-03-07 13:40:13,747 INFO recall_rcnn_0.3: 0.936902 2024-03-07 13:40:13,747 INFO recall_roi_0.5: 0.000000 2024-03-07 13:40:13,747 INFO recall_rcnn_0.5: 0.875569 2024-03-07 13:40:13,747 INFO recall_roi_0.7: 0.000000 2024-03-07 13:40:13,747 INFO recall_rcnn_0.7: 0.642027 2024-03-07 13:40:13,749 INFO Average predicted number of objects(3769 samples): 15.823 2024-03-07 13:40:30,185 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7899, 89.5276, 88.3655 bev AP:89.8089, 86.8546, 84.7272 3d AP:87.5301, 77.4915, 75.6279 aos AP:90.77, 89.36, 88.06 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.5618, 91.9613, 90.8331 bev AP:92.0819, 87.7806, 86.5679 3d AP:88.3309, 78.7233, 75.6268 aos AP:95.54, 91.77, 90.50 Car AP@0.70, 0.50, 0.50: bbox AP:90.7899, 89.5276, 88.3655 bev AP:90.7887, 89.9586, 89.0858 3d AP:90.7887, 89.8713, 88.9199 aos AP:90.77, 89.36, 88.06 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.5618, 91.9613, 90.8331 bev AP:95.5930, 94.4635, 93.5686 3d AP:95.5759, 94.1879, 92.9782 aos AP:95.54, 91.77, 90.50 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:68.2071, 64.5808, 60.8431 bev AP:57.7600, 53.3384, 48.9331 3d AP:52.9999, 47.6337, 43.4523 aos AP:63.94, 59.81, 55.87 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:68.6781, 64.6337, 60.4548 bev AP:57.8328, 52.1277, 47.3334 3d AP:51.5507, 46.1071, 41.0419 aos AP:63.87, 59.34, 54.97 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:68.2071, 64.5808, 60.8431 bev AP:74.1267, 70.4830, 66.6133 3d AP:73.9275, 70.2833, 66.3509 aos AP:63.94, 59.81, 55.87 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:68.6781, 64.6337, 60.4548 bev AP:74.9390, 71.3683, 67.0309 3d AP:74.7165, 71.1119, 66.6024 aos AP:63.87, 59.34, 54.97 Cyclist AP@0.50, 0.50, 0.50: bbox AP:86.6789, 78.4154, 74.0690 bev AP:82.9717, 71.7719, 67.6408 3d AP:82.0319, 69.2462, 64.5875 aos AP:86.39, 77.54, 73.20 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:91.0006, 79.8313, 75.3468 bev AP:86.6763, 72.4468, 67.4855 3d AP:84.2363, 69.9676, 65.1350 aos AP:90.65, 78.83, 74.41 Cyclist AP@0.50, 0.25, 0.25: bbox AP:86.6789, 78.4154, 74.0690 bev AP:90.9554, 76.5323, 72.8491 3d AP:90.9554, 76.5323, 72.8491 aos AP:86.39, 77.54, 73.20 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:91.0006, 79.8313, 75.3468 bev AP:91.5256, 77.5414, 73.3682 3d AP:91.5256, 77.5414, 73.3682 aos AP:90.65, 78.83, 74.41

2024-03-07 13:40:30,187 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_77/val 2024-03-07 13:40:30,187 INFO ****Evaluation done.** 2024-03-07 13:40:30,207 INFO Epoch 77 has been evaluated 2024-03-07 13:40:30,208 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_78.pth to CPU 2024-03-07 13:40:30,236 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:40:30,552 INFO ==> Done (loaded 163/163) 2024-03-07 13:40:30,554 INFO EPOCH 78 EVALUATION ** 2024-03-07 13:41:48,706 INFO Performance of EPOCH 78 ***** 2024-03-07 13:41:48,706 INFO Generate label finished(sec_per_example: 0.0207 second). 2024-03-07 13:41:48,707 INFO recall_roi_0.3: 0.000000 2024-03-07 13:41:48,707 INFO recall_rcnn_0.3: 0.936902 2024-03-07 13:41:48,707 INFO recall_roi_0.5: 0.000000 2024-03-07 13:41:48,707 INFO recall_rcnn_0.5: 0.875114 2024-03-07 13:41:48,707 INFO recall_roi_0.7: 0.000000 2024-03-07 13:41:48,707 INFO recall_rcnn_0.7: 0.642084 2024-03-07 13:41:48,708 INFO Average predicted number of objects(3769 samples): 15.655 2024-03-07 13:42:05,220 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7756, 89.5155, 88.3488 bev AP:89.7704, 86.6546, 84.1411 3d AP:87.5078, 77.4915, 75.5705 aos AP:90.76, 89.33, 88.03 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.5645, 91.9839, 90.8331 bev AP:92.0782, 87.7187, 86.3860 3d AP:88.3449, 78.7161, 75.6452 aos AP:95.55, 91.78, 90.49 Car AP@0.70, 0.50, 0.50: bbox AP:90.7756, 89.5155, 88.3488 bev AP:90.7744, 89.9268, 89.0527 3d AP:90.7744, 89.8419, 88.8822 aos AP:90.76, 89.33, 88.03 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.5645, 91.9839, 90.8331 bev AP:95.5947, 94.4154, 93.5558 3d AP:95.5777, 94.1155, 92.8494 aos AP:95.55, 91.78, 90.49 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:68.8301, 64.8144, 61.0087 bev AP:58.0412, 53.5972, 48.9339 3d AP:52.9867, 47.7790, 43.4490 aos AP:64.73, 60.26, 56.27 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:69.4347, 64.8942, 60.6928 bev AP:57.8916, 52.2187, 47.3725 3d AP:51.8554, 46.0167, 41.0737 aos AP:64.86, 59.76, 55.46 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:68.8301, 64.8144, 61.0087 bev AP:74.5291, 70.7279, 66.7222 3d AP:74.3342, 70.3765, 66.4881 aos AP:64.73, 60.26, 56.27 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:69.4347, 64.8942, 60.6928 bev AP:75.1063, 71.5706, 67.0829 3d AP:74.8853, 71.0367, 66.6919 aos AP:64.86, 59.76, 55.46 Cyclist AP@0.50, 0.50, 0.50: bbox AP:86.8568, 78.8440, 75.1784 bev AP:82.8023, 71.8662, 67.9060 3d AP:81.9468, 69.1357, 64.5864 aos AP:86.57, 78.07, 74.40 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:91.1494, 80.0259, 75.9558 bev AP:85.2334, 72.5377, 67.6096 3d AP:84.2081, 70.0129, 65.0231 aos AP:90.78, 79.15, 75.10 Cyclist AP@0.50, 0.25, 0.25: bbox AP:86.8568, 78.8440, 75.1784 bev AP:90.9229, 76.6633, 73.2120 3d AP:90.9229, 76.6633, 73.2120 aos AP:86.57, 78.07, 74.40 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:91.1494, 80.0259, 75.9558 bev AP:91.5942, 77.6813, 73.7049 3d AP:91.5942, 77.6813, 73.7049 aos AP:90.78, 79.15, 75.10

2024-03-07 13:42:05,225 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_78/val 2024-03-07 13:42:05,225 INFO ****Evaluation done.** 2024-03-07 13:42:05,245 INFO Epoch 78 has been evaluated 2024-03-07 13:42:05,246 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_79.pth to CPU 2024-03-07 13:42:05,273 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:42:05,587 INFO ==> Done (loaded 163/163) 2024-03-07 13:42:05,589 INFO EPOCH 79 EVALUATION ** 2024-03-07 13:43:23,683 INFO Performance of EPOCH 79 ***** 2024-03-07 13:43:23,684 INFO Generate label finished(sec_per_example: 0.0207 second). 2024-03-07 13:43:23,684 INFO recall_roi_0.3: 0.000000 2024-03-07 13:43:23,684 INFO recall_rcnn_0.3: 0.936674 2024-03-07 13:43:23,684 INFO recall_roi_0.5: 0.000000 2024-03-07 13:43:23,684 INFO recall_rcnn_0.5: 0.875000 2024-03-07 13:43:23,684 INFO recall_roi_0.7: 0.000000 2024-03-07 13:43:23,684 INFO recall_rcnn_0.7: 0.642426 2024-03-07 13:43:23,686 INFO Average predicted number of objects(3769 samples): 15.754 2024-03-07 13:43:40,021 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7756, 89.5124, 88.3615 bev AP:89.7775, 86.7088, 84.4590 3d AP:87.4881, 77.4619, 75.6121 aos AP:90.76, 89.33, 88.05 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.5045, 91.9229, 90.7980 bev AP:92.0176, 87.6988, 86.4492 3d AP:88.2893, 78.6600, 75.6160 aos AP:95.49, 91.73, 90.45 Car AP@0.70, 0.50, 0.50: bbox AP:90.7756, 89.5124, 88.3615 bev AP:90.7744, 89.9326, 89.0511 3d AP:90.7744, 89.8554, 88.8830 aos AP:90.76, 89.33, 88.05 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.5045, 91.9229, 90.7980 bev AP:95.5305, 94.3727, 93.5182 3d AP:95.5144, 94.0833, 92.8413 aos AP:95.49, 91.73, 90.45 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:68.5497, 64.6122, 60.9149 bev AP:57.3171, 53.4534, 48.7458 3d AP:52.6536, 47.4594, 43.2900 aos AP:64.41, 60.09, 56.18 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:69.0288, 64.6470, 60.5038 bev AP:57.4332, 52.0594, 47.1590 3d AP:51.2586, 45.7375, 40.8415 aos AP:64.48, 59.57, 55.29 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:68.5497, 64.6122, 60.9149 bev AP:74.3007, 70.7014, 66.6368 3d AP:74.0922, 70.3307, 66.3322 aos AP:64.41, 60.09, 56.18 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:69.0288, 64.6470, 60.5038 bev AP:74.9962, 71.4462, 67.0466 3d AP:74.7608, 70.8949, 66.5598 aos AP:64.48, 59.57, 55.29 Cyclist AP@0.50, 0.50, 0.50: bbox AP:86.8878, 78.7610, 75.1054 bev AP:82.8284, 71.6790, 67.8402 3d AP:82.1074, 69.2402, 64.5038 aos AP:86.60, 77.85, 74.21 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:91.2340, 79.9053, 75.8184 bev AP:85.3194, 72.4648, 67.5670 3d AP:84.4631, 69.9697, 65.0966 aos AP:90.86, 78.90, 74.84 Cyclist AP@0.50, 0.25, 0.25: bbox AP:86.8878, 78.7610, 75.1054 bev AP:91.0495, 76.8781, 72.8392 3d AP:91.0495, 76.8781, 72.8392 aos AP:86.60, 77.85, 74.21 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:91.2340, 79.9053, 75.8184 bev AP:91.6782, 77.6336, 73.4780 3d AP:91.6782, 77.6336, 73.4780 aos AP:90.86, 78.90, 74.84

2024-03-07 13:43:40,026 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_79/val 2024-03-07 13:43:40,026 INFO ****Evaluation done.** 2024-03-07 13:43:40,044 INFO Epoch 79 has been evaluated 2024-03-07 13:43:40,045 INFO ==> Loading parameters from checkpoint /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/ckpt/checkpoint_epoch_80.pth to CPU 2024-03-07 13:43:40,071 INFO ==> Checkpoint trained from version: pcdet+0.6.0+255db8f+pybd06a67 2024-03-07 13:43:40,198 INFO ==> Done (loaded 163/163) 2024-03-07 13:43:40,199 INFO EPOCH 80 EVALUATION ** 2024-03-07 13:44:57,745 INFO Performance of EPOCH 80 ***** 2024-03-07 13:44:57,745 INFO Generate label finished(sec_per_example: 0.0206 second). 2024-03-07 13:44:57,745 INFO recall_roi_0.3: 0.000000 2024-03-07 13:44:57,745 INFO recall_rcnn_0.3: 0.936788 2024-03-07 13:44:57,745 INFO recall_roi_0.5: 0.000000 2024-03-07 13:44:57,745 INFO recall_rcnn_0.5: 0.874886 2024-03-07 13:44:57,745 INFO recall_roi_0.7: 0.000000 2024-03-07 13:44:57,745 INFO recall_rcnn_0.7: 0.642255 2024-03-07 13:44:57,747 INFO Average predicted number of objects(3769 samples): 15.767 2024-03-07 13:45:14,842 INFO Car AP@0.70, 0.70, 0.70: bbox AP:90.7756, 89.5129, 88.3751 bev AP:89.7809, 86.7513, 84.5777 3d AP:87.5187, 77.4911, 75.5956 aos AP:90.76, 89.34, 88.06 Car AP_R40@0.70, 0.70, 0.70: bbox AP:95.5184, 91.9351, 90.7985 bev AP:92.0357, 87.7109, 86.4892 3d AP:88.3079, 78.6775, 75.6115 aos AP:95.50, 91.74, 90.45 Car AP@0.70, 0.50, 0.50: bbox AP:90.7756, 89.5129, 88.3751 bev AP:90.7710, 89.9168, 89.0549 3d AP:90.7710, 89.8515, 88.8665 aos AP:90.76, 89.34, 88.06 Car AP_R40@0.70, 0.50, 0.50: bbox AP:95.5184, 91.9351, 90.7985 bev AP:95.5397, 94.3762, 93.5158 3d AP:95.5235, 94.0902, 92.7921 aos AP:95.50, 91.74, 90.45 Pedestrian AP@0.50, 0.50, 0.50: bbox AP:68.7350, 64.8406, 61.1076 bev AP:57.4579, 53.5965, 48.7089 3d AP:52.8531, 47.6276, 43.3660 aos AP:64.59, 60.20, 56.33 Pedestrian AP_R40@0.50, 0.50, 0.50: bbox AP:69.3595, 64.9338, 60.7355 bev AP:57.5283, 52.2262, 47.2076 3d AP:51.3809, 45.8878, 40.9421 aos AP:64.76, 59.73, 55.44 Pedestrian AP@0.50, 0.25, 0.25: bbox AP:68.7350, 64.8406, 61.1076 bev AP:74.5357, 70.7575, 66.6754 3d AP:74.3363, 70.4050, 66.4326 aos AP:64.59, 60.20, 56.33 Pedestrian AP_R40@0.50, 0.25, 0.25: bbox AP:69.3595, 64.9338, 60.7355 bev AP:75.1337, 71.5848, 67.1199 3d AP:74.9109, 71.0166, 66.7015 aos AP:64.76, 59.73, 55.44 Cyclist AP@0.50, 0.50, 0.50: bbox AP:86.8128, 78.7895, 75.1452 bev AP:82.7761, 71.7918, 67.8346 3d AP:82.0577, 69.2279, 64.5570 aos AP:86.53, 78.01, 74.36 Cyclist AP_R40@0.50, 0.50, 0.50: bbox AP:91.2360, 79.9652, 75.9099 bev AP:85.2570, 72.5337, 67.5499 3d AP:84.4018, 70.0027, 65.1040 aos AP:90.86, 79.08, 75.05 Cyclist AP@0.50, 0.25, 0.25: bbox AP:86.8128, 78.7895, 75.1452 bev AP:91.0714, 76.6888, 72.8220 3d AP:91.0714, 76.6888, 72.8220 aos AP:86.53, 78.01, 74.36 Cyclist AP_R40@0.50, 0.25, 0.25: bbox AP:91.2360, 79.9652, 75.9099 bev AP:91.6310, 77.6228, 73.5715 3d AP:91.6310, 77.6228, 73.5715 aos AP:90.86, 79.08, 75.05

2024-03-07 13:45:14,843 INFO Result is saved to /hdd_10tb/sina/Radial_MAE/output/kitti_models/second/default/eval/eval_with_train/epoch_80/val 2024-03-07 13:45:14,844 INFO ****Evaluation done.***** 2024-03-07 13:45:14,864 INFO Epoch 80 has been evaluated 2024-03-07 13:45:44,895 INFO **End evaluation kitti_models/second(default)**

sinatayebati commented 4 months ago

It's also worth noting that when I evaluate using ckpt published by OpenPCDet, I get the exact same accuracy as reported. Thus, I assume it's a problem of data infos. This problem only happens when I train from scratch.

Another observation is that when I train any of models from scratch on other datasets including Waymo and NuScene, I can reproduce the same exact accuracy. Thus, this observation also tells me it's safe to assume the environment is not an issue as well.

I'm still struggling with this issue and cannot find the root cause why I cannot reproduce your reported results by training from scratch on KITTI, so any hints or suggestions would be greatly appreciated.

Regarding my environment, I'm using pytorch=1.10.2, spconv=2.3.6, and the rest is attached bellow:

name: rd-mae channels:

github-actions[bot] commented 3 months ago

This issue is stale because it has been open for 30 days with no activity.

github-actions[bot] commented 3 months ago

This issue was closed because it has been inactive for 14 days since being marked as stale.

zhangye1111 commented 1 month ago

还值得注意的是，当我使用 OpenPCDet 发布的 ckpt 进行评估时，我得到的准确率与报告的完全相同。因此，我认为这是数据信息的问题。这个问题只发生在我从头开始训练的时候。

另一个观察结果是，当我在其他数据集（包括 Waymo 和 NuScene）上从头训练任何模型时，我都可以重现完全相同的准确率。因此，这个观察结果也告诉我，可以安全地假设环境也不是问题。

我仍在努力解决这个问题，但找不到导致我无法通过在 KITTI 上从头开始训练来重现您报告的结果的根本原因，因此，任何提示或建议都将不胜感激。

关于我的环境，我使用的是 pytorch=1.10.2、spconv=2.3.6，其余部分附在下面：

名称：rd-mae 频道：

  • pytorch3d
  • 阴谋地
  • 瓶装商
  • 免疫学
  • 核心数
  • pythorch
  • 康达锻造公司
  • 默认 依赖项：
  • cudatoolkit-dev=11.3.1
  • 绘图=5.8.1
  • plotly_express=0.4.1
  • 蟒蛇=3.7.13
  • 火炬=1.10.2
  • pytorch3d=0.6.2
  • pyyaml=6.0
  • 设置工具=62.3.3
  • torchvision=0.11.3
  • tqdm=4.64.0
  • 点子
  • 内核
  • Python

  • 点子：

    • argparse==1.4.0
    • gdown==4.4.0
    • h5py==3.7.0
    • 易词典
    • matplotlib==3.5.2
    • matplotlib-inline==0.1.3
    • open3d==0.15.2
    • opencv-python==4.6.0.66
    • 熊猫==1.3.5
    • tensorboardx==2.5.1
    • timm==0.4.5
    • tinycss2==1.1.1
    • tomlkit==0.11.0
    • traitlets==5.2.2.post1
    • transforms3d==0.3.1
    • 忍者
    • faiss-cpu
    • 万德布

May I ask if you have solved this problem? I get low pedestrian detection as well.