hv_pointpillars_secfpn_6x8_160e_kitti-3d-3class.py

_base_ = [
    '../_base_/models/hv_pointpillars_secfpn_kitti.py',
    '../_base_/datasets/kitti-3d-3class.py',
    '../_base_/schedules/cyclic_40e.py', '../_base_/default_runtime.py'
]

point_cloud_range = [0, -39.68, -3, 69.12, 39.68, 1]
# dataset settings
data_root = 'data/kitti/'
class_names = ['Pedestrian', 'Cyclist', 'Car']
metainfo = dict(CLASSES=class_names)
# PointPillars adopted a different sampling strategies among classes
db_sampler = dict(
    data_root=data_root,
    info_path=data_root + 'kitti_dbinfos_train.pkl',
    rate=1.0,
    prepare=dict(
        filter_by_difficulty=[-1],
        filter_by_min_points=dict(Car=5, Pedestrian=5, Cyclist=5)),
    classes=class_names,
    sample_groups=dict(Car=15, Pedestrian=15, Cyclist=15))

# PointPillars uses different augmentation hyper parameters
train_pipeline = [
    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=4, use_dim=4),
    dict(type='LoadAnnotations3D', with_bbox_3d=True, with_label_3d=True),
    dict(type='ObjectSample', db_sampler=db_sampler, use_ground_plane=True),
    dict(type='RandomFlip3D', flip_ratio_bev_horizontal=0.5),
    dict(
        type='GlobalRotScaleTrans',
        rot_range=[-0.78539816, 0.78539816],
        scale_ratio_range=[0.95, 1.05]),
    dict(type='PointsRangeFilter', point_cloud_range=point_cloud_range),
    dict(type='ObjectRangeFilter', point_cloud_range=point_cloud_range),
    dict(type='PointShuffle'),
    dict(
        type='Pack3DDetInputs',
        keys=['points', 'gt_labels_3d', 'gt_bboxes_3d'])
]
test_pipeline = [
    dict(type='LoadPointsFromFile', coord_type='LIDAR', load_dim=4, use_dim=4),
    dict(
        type='MultiScaleFlipAug3D',
        img_scale=(1333, 800),
        pts_scale_ratio=1,
        flip=False,
        transforms=[
            dict(
                type='GlobalRotScaleTrans',
                rot_range=[0, 0],
                scale_ratio_range=[1., 1.],
                translation_std=[0, 0, 0]),
            dict(type='RandomFlip3D'),
            dict(
                type='PointsRangeFilter', point_cloud_range=point_cloud_range),
            dict(type='Pack3DDetInputs', keys=['points'])
        ])
]

train_dataloader = dict(
    dataset=dict(dataset=dict(pipeline=train_pipeline, metainfo=metainfo)))
test_dataloader = dict(dataset=dict(metainfo=metainfo))
val_dataloader = dict(dataset=dict(metainfo=metainfo))
# In practice PointPillars also uses a different schedule
# optimizer
lr = 0.001
epoch_num = 80
iter_num_in_epoch = 3712
optim_wrapper = dict(
    optimizer=dict(lr=lr), clip_grad=dict(max_norm=35, norm_type=2))
param_scheduler = [
    dict(
        type='CosineAnnealingLR',
        T_max=epoch_num * 0.4 * iter_num_in_epoch,
        eta_min=lr * 10,
        by_epoch=False,
        begin=0,
        end=epoch_num * 0.4 * iter_num_in_epoch),
    dict(
        type='CosineAnnealingLR',
        T_max=epoch_num * 0.6 * iter_num_in_epoch,
        eta_min=lr * 1e-4,
        by_epoch=False,
        begin=epoch_num * 0.4 * iter_num_in_epoch,
        end=epoch_num * 1 * iter_num_in_epoch),
    dict(
        type='CosineAnnealingBetas',
        T_max=epoch_num * 0.4 * iter_num_in_epoch,
        eta_min=0.85 / 0.95,
        by_epoch=False,
        begin=0,
        end=epoch_num * 0.4 * iter_num_in_epoch),
    dict(
        type='CosineAnnealingBetas',
        T_max=epoch_num * 0.6 * iter_num_in_epoch,
        eta_min=1,
        by_epoch=False,
        begin=epoch_num * 0.4 * iter_num_in_epoch,
        end=epoch_num * 1 * iter_num_in_epoch)
]
# max_norm=35 is slightly better than 10 for PointPillars in the earlier
# development of the codebase thus we keep the setting. But we does not
# specifically tune this parameter.
# PointPillars usually need longer schedule than second, we simply double
# the training schedule. Do remind that since we use RepeatDataset and
# repeat factor is 2, so we actually train 160 epochs.
train_cfg = dict(by_epoch=True, max_epochs=epoch_num)
val_cfg = dict(interval=2)
test_cfg = dict()