sparsefusion_nusc_voxel_LC_r50.py

point_cloud_range = [-54.0, -54.0, -5.0, 54.0, 54.0, 3.0]
class_names = [
    'car', 'truck', 'construction_vehicle', 'bus', 'trailer', 'barrier',
    'motorcycle', 'bicycle', 'pedestrian', 'traffic_cone'
]
voxel_size = [0.075, 0.075, 0.2]
out_size_factor = 8
evaluation = dict(interval=1)
dataset_type = 'NuScenesDataset_ViewInfo'
data_root = 'data/nuscenes/'
input_modality = dict(
    use_lidar=True,
    use_camera=True,
    use_radar=False,
    use_map=False,
    use_external=False)
img_scale = (800, 448)
num_views = 6
img_norm_cfg = dict(mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
train_pipeline = [
    dict(
        type='LoadPointsFromFile',
        coord_type='LIDAR',
        load_dim=5,
        use_dim=[0, 1, 2, 3, 4],
    ),
    dict(
        type='LoadPointsFromMultiSweeps',
        sweeps_num=10,
        use_dim=[0, 1, 2, 3, 4],
    ),
    dict(type='MyLoadAnnotations3D', with_bbox_3d=True, with_label_3d=True, with_bbox=True, with_label=True, with_centers=True, with_cam_bbox=True, with_visible=True),
    dict(type='LoadMultiViewImageFromFiles'),
    dict(
        type='OurGlobalRotScaleTrans',
        rot_range=[-0.3925 * 2, 0.3925 * 2],
        scale_ratio_range=[0.9, 1.1],
        translation_std=[0.5, 0.5, 0.5],
    ),
    dict(
        type='OurRandomFlip3D',
        sync_2d=False,
        flip_ratio_bev_horizontal=0.5,
        flip_ratio_bev_vertical=0.5),
    # dict(type='PhotoMetricDistortionMultiViewImage', swap_channel=False),  # color augmentation cannot improve the performance
    dict(type='OurRandomAffine', scaling_ratio_range=(0.9, 1.1), flip_ratio=0.5, flip_sync_3d=True),
    dict(type='MyResize', img_scale=img_scale, keep_ratio=True),
    dict(type='MyNormalize', **img_norm_cfg),
    dict(type='MyPad', size_divisor=32),
    dict(type='SparseDepth', scale_factors=[4], exp_time=0),
    dict(type='PointsRangeFilter', point_cloud_range=point_cloud_range),
    dict(type='OurObjectRangeFilter', point_cloud_range=point_cloud_range),
    dict(type='ObjectNameFilter', classes=class_names),
    dict(type='PointShuffle'),
    dict(type='DefaultFormatBundle3D', class_names=class_names),
    dict(type='Collect3D', keys=['points', 'img', 'gt_bboxes_3d', 'gt_labels_3d', 'gt_bboxes', 'gt_labels', 'gt_pts_centers_view', 'gt_img_centers_view', 'gt_bboxes_cam_view', 'gt_bboxes_lidar_view', 'sparse_depth', 'gt_visible_3d'])
]
test_pipeline = [
    dict(
        type='LoadPointsFromFile',
        coord_type='LIDAR',
        load_dim=5,
        use_dim=[0, 1, 2, 3, 4],
    ),
    dict(
        type='LoadPointsFromMultiSweeps',
        sweeps_num=10,
        use_dim=[0, 1, 2, 3, 4],
    ),
    dict(type='LoadMultiViewImageFromFiles'),
    dict(
        type='MultiScaleFlipAug3D',
        img_scale=img_scale,
        pts_scale_ratio=1,
        flip=False,
        transforms=[
            dict(
                type='GlobalRotScaleTrans',
                rot_range=[0, 0],
                scale_ratio_range=[1.0, 1.0],
                translation_std=[0, 0, 0]),
            dict(type='RandomFlip3D'),
            dict(type='MyResize', img_scale=img_scale, keep_ratio=True),
            dict(type='MyNormalize', **img_norm_cfg),
            dict(type='MyPad', size_divisor=32),
            dict(type='SparseDepth', scale_factors=[4]),
            dict(
                type='DefaultFormatBundle3D',
                class_names=class_names,
                with_label=False),
            dict(type='Collect3D', keys=['points', 'img', 'sparse_depth'])
        ])
]

# our default setting uses 4 GPUs with 4 samples per-GPU, please ensure the LR consistent with your batch size
data = dict(
    samples_per_gpu=4,
    workers_per_gpu=4,
    train=dict(
        type='CBGSDataset',
        dataset=dict(
            type=dataset_type,
            data_root=data_root,
            num_views=num_views,
            ann_file=data_root + '/nuscenes_infos_w_views_train.pkl',
            load_interval=1,
            pipeline=train_pipeline,
            classes=class_names,
            modality=input_modality,
            test_mode=False,
            box_type_3d='LiDAR')),
    val=dict(
        type=dataset_type,
        data_root=data_root,
        num_views=num_views,
        ann_file=data_root + '/nuscenes_infos_w_views_val.pkl',
        load_interval=1,
        pipeline=test_pipeline,
        classes=class_names,
        modality=input_modality,
        test_mode=True,
        box_type_3d='LiDAR'),
    test=dict(
        type=dataset_type,
        data_root=data_root,
        num_views=num_views,
        ann_file=data_root + '/nuscenes_infos_w_views_val.pkl',
        load_interval=1,
        pipeline=test_pipeline,
        classes=class_names,
        modality=input_modality,
        test_mode=True,
        box_type_3d='LiDAR'))
model = dict(
    type='SparseFusionDetector',
    freeze_img=False,
    img_backbone=dict(
        type='ResNet',
        depth=50,
        num_stages=4,
        out_indices=(0, 1, 2, 3),
        frozen_stages=1,
        in_channels=3,
        norm_cfg=dict(type='BN', requires_grad=True),
        norm_eval=True,
        style='pytorch',
    ),
    img_neck=dict(
        type='FPN',
        in_channels=[256, 512, 1024, 2048],
        out_channels=256,
        num_outs=5),
    pts_voxel_layer=dict(
        max_num_points=10,
        voxel_size=voxel_size,
        max_voxels=(120000, 160000),
        point_cloud_range=point_cloud_range),
    pts_voxel_encoder=dict(
        type='HardSimpleVFE',
        num_features=5,
    ),
    pts_middle_encoder=dict(
        type='SparseEncoder',
        in_channels=5,
        sparse_shape=[41, 1440, 1440],
        output_channels=128,
        order=('conv', 'norm', 'act'),
        encoder_channels=((16, 16, 32), (32, 32, 64), (64, 64, 128), (128, 128)),
        encoder_paddings=((0, 0, 1), (0, 0, 1), (0, 0, [0, 1, 1]), (0, 0)),
        block_type='basicblock'),
    pts_backbone=dict(
        type='SECOND',
        in_channels=256,
        out_channels=[128, 256],
        layer_nums=[5, 5],
        layer_strides=[1, 2],
        norm_cfg=dict(type='BN', eps=0.001, momentum=0.01),
        conv_cfg=dict(type='Conv2d', bias=False)),
    pts_neck=dict(
        type='SECONDFPN',
        in_channels=[128, 256],
        out_channels=[256, 256],
        upsample_strides=[1, 2],
        norm_cfg=dict(type='BN', eps=0.001, momentum=0.01),
        upsample_cfg=dict(type='deconv', bias=False),
        use_conv_for_no_stride=True),
    pts_bbox_head=dict(
        type='SparseFusionHead2D_Deform',
        num_views=num_views,
        in_channels_img=256,
        out_size_factor_img=4,
        in_channels=256 * 2,
        hidden_channel=128,
        num_heads=8,
        num_classes=len(class_names),
        ffn_channel=256,
        dropout=0.1,
        bn_momentum=0.1,
        activation='relu',
        img_reg_bn=False,
        img_reg_layer=3,
        common_heads=dict(center=(2, 2), height=(1, 2), dim=(3, 2), rot=(2, 2), vel=(2, 2)),

        num_proposals=200,  # query number in the LiDAR branch
        num_img_proposals=200,  # query number in the camera branch

        level_num=4,

        num_pts_decoder_layers=1,  # number of transformer layers in the point detector (if you set it >1, ensure it is consistent with your pretrained LiDAR-only model or set "freeze_lidar_detector=False")
        num_img_decoder_layers=1,  # number of transformer layers in the image detector
        num_fusion_decoder_layers=1,  # number of the transformer layers in the fusion stage

        initialize_by_heatmap=True,  # initialize the queries based on the heatmap (we never set it as False)

        semantic_transfer=True,  # whether to use semantic transfer (camera to LiDAR)
        cross_only=True,  # if false, output heatmap would be the average of semantic transfer and the LiDAR-only heatmap of TransFusion-L
        cross_heatmap_layer=1,
        nms_kernel_size=3,  # suppress nearby proposals when initializing queries for the LiDAR branch

        geometric_transfer=True,  # whether to use geometric transfer
        depth_input_channel=2,  # channel number of depth features. Do not change it unless you modify the SparseDepth class in "mmdet3d/datasets/pipelines/loading.py"
        img_heatmap_layer=2,
        img_nms_kernel_size=3, # suppress nearby proposals when initializing queries for the camera branch

        view_transform=True,  # whether to transform the coordinate for the output bboxes of the camera branch
        use_camera='se',  # "se" or None: whether to encode the camera parameters in the view transformation

        bbox_coder=dict(
            type='TransFusionBBoxCoder',
            pc_range=point_cloud_range[:2],
            voxel_size=voxel_size[:2],
            out_size_factor=out_size_factor,
            post_center_range=[-61.2, -61.2, -10.0, 61.2, 61.2, 10.0],
            score_threshold=0.0,
            code_size=10,
        ),
        bbox_2d_coder=dict(
            type='CameraBBoxCoder',
            code_size=10,
        ),
        loss_cls=dict(type='FocalLoss', use_sigmoid=True, gamma=2, alpha=0.25, reduction='mean', loss_weight=1.0),
        loss_bbox=dict(type='L1Loss', reduction='mean', loss_weight=0.25),
        loss_heatmap=dict(type='GaussianFocalLoss', reduction='mean', loss_weight=0.1),
        loss_heatmap_2d=dict(type='GaussianFocalLoss', reduction='mean', loss_weight=0.1),
        loss_center_2d=dict(type='L1Loss', reduction='mean', loss_weight=5.0),
    ),
    train_cfg=dict(
        pts=dict(
            dataset='nuScenes',
            assigner=dict(
                type='HungarianAssigner3D',
                iou_calculator=dict(type='BboxOverlaps3D', coordinate='lidar'),
                cls_cost=dict(type='FocalLossCost', gamma=2, alpha=0.25, weight=0.15),
                reg_cost=dict(type='BBoxBEVL1Cost', weight=0.25),
                iou_cost=dict(type='IoU3DCost', weight=0.25)
            ),
            assigner_2d=dict(
                type='HungarianAssignerCameraBox',
                iou_calculator=dict(type='BboxOverlaps3D', coordinate='camera'),
                cls_cost=dict(type='FocalLossCost', gamma=2, alpha=0.25, weight=0.15),
                reg_cost=dict(type='BBoxBEVL1Cost', weight=0.25),
                iou_cost=dict(type='IoU3DCost', weight=0.25),
            ),
            pos_weight=-1,
            gaussian_overlap=0.1,
            gaussian_overlap_2d=0.1,
            min_radius=2,
            max_radius=999,
            grid_size=[1440, 1440, 40],  # [x_len, y_len, 1]
            voxel_size=voxel_size,
            out_size_factor=out_size_factor,
            code_weights=[1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.2, 0.2],
            img_code_weights=[1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.2, 0.2],
            point_cloud_range=point_cloud_range)),
    test_cfg=dict(
        pts=dict(
            dataset='nuScenes',
            grid_size=[1440, 1440, 40],
            img_scale=img_scale,
            out_size_factor=out_size_factor,
            pc_range=point_cloud_range,
            voxel_size=voxel_size,
            nms_type='circle',
        )))
optimizer = dict(
    type='AdamW',
    lr=0.0001,
    weight_decay=0.01,
    paramwise_cfg=dict(
        custom_keys={
            'img_backbone': dict(lr_mult=0.1),
            'img_neck': dict(lr_mult=0.1),
            'pts_voxel_layer': dict(lr_mult=0.1),
            'pts_voxel_encoder': dict(lr_mult=0.1),
            'pts_middle_encoder': dict(lr_mult=0.1),
            'pts_backbone': dict(lr_mult=0.1),
            'pts_neck': dict(lr_mult=0.1),
            'pts_bbox_head.point_transformer': dict(lr_mult=0.1),
            'pts_bbox_head.class_encoding': dict(lr_mult=0.1),
            'pts_bbox_head.heatmap_head': dict(lr_mult=0.1),
            'pts_bbox_head.shared_conv': dict(lr_mult=0.1),
        }),
)  # for 4gpu * 4sample_per_gpu
optimizer_config = dict(grad_clip=dict(max_norm=0.1, norm_type=2))

lr_config = dict(
    policy='cyclic',
    target_ratio=(8, 0.0001),
    cyclic_times=1,
    step_ratio_up=0.4,
    )
momentum_config = dict(
    policy='cyclic',
    target_ratio=(0.8947368421052632, 1),
    cyclic_times=1,
    step_ratio_up=0.4)
total_epochs = 6
checkpoint_config = dict(interval=1)
log_config = dict(
    interval=50,
    hooks=[dict(type='TextLoggerHook'),
           dict(type='TensorboardLoggerHook')])
dist_params = dict(backend='nccl')
log_level = 'INFO'
work_dir = None
load_from = 'checkpoints/sparsefusion_voxel0075_R50_initial.pth'
resume_from = None
workflow = [('train', 1)]
gpu_ids = range(0, 8)

freeze_lidar_components = True  # freeze the LiDAR backbone
freeze_lidar_detector = True  # freeze the LiDAR detector

find_unused_parameters = True


# Evaluating bboxes of pts_bbox
# mAP: 0.7051
# mATE: 0.2757
# mASE: 0.2506
# mAOE: 0.2767
# mAVE: 0.2562
# mAAE: 0.1869
# NDS: 0.7279
# Eval time: 137.2s
#
# Per-class results:
# Object Class    AP      ATE     ASE     AOE     AVE     AAE
# car     0.883   0.171   0.146   0.066   0.262   0.187
# truck   0.643   0.305   0.177   0.071   0.235   0.211
# bus     0.775   0.304   0.177   0.044   0.411   0.250
# trailer 0.447   0.522   0.214   0.432   0.179   0.159
# construction_vehicle    0.303   0.669   0.424   0.842   0.127   0.326
# pedestrian      0.898   0.127   0.282   0.329   0.216   0.104
# motorcycle      0.810   0.189   0.241   0.215   0.426   0.249
# bicycle 0.712   0.164   0.263   0.422   0.193   0.010
# traffic_cone    0.808   0.118   0.309   nan     nan     nan
# barrier 0.772   0.188   0.273   0.068   nan     nan