rtmpose-m_8xb256-420e-512x512.py

default_scope = 'mmpose'

# multiprocessing backend
env_cfg = dict(
    cudnn_benchmark=False,
    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
    dist_cfg=dict(backend='nccl'),
)

# visualizer
vis_backends = [dict(type='LocalVisBackend'),]
visualizer = dict(
    type='PoseLocalVisualizer',
    vis_backends=vis_backends,
    name='visualizer'
)

# logger
log_processor = dict(type='LogProcessor', window_size=50, by_epoch=True, num_digits=6)
log_level = 'INFO'
load_from = None
resume = False

# file I/O backend
backend_args = dict(backend='local')

# runtime
max_epochs = 420
stage2_num_epochs = 20  # when to start stage 2 training
base_lr = 5e-3

randomness = dict(seed=2023)

# training/validation/testing progress
train_cfg = dict(by_epoch=True, max_epochs=max_epochs, val_interval=2)
val_cfg = dict()
test_cfg = dict()

# optimizer
optim_wrapper = dict(
    type='OptimWrapper',
    optimizer=dict(type='AdamW', lr=base_lr, weight_decay=0.05),
    paramwise_cfg=dict(
        norm_decay_mult=0, bias_decay_mult=0, bypass_duplicate=True))

# learning rate
param_scheduler = [
    dict(
        type='LinearLR',
        start_factor=1.0e-5,
        by_epoch=False,
        begin=0,
        end=1000),
    dict(
        # use cosine lr from 210 to 420 epoch
        type='CosineAnnealingLR',
        eta_min=base_lr * 0.05,
        begin=max_epochs // 2,
        end=max_epochs,
        T_max=max_epochs // 2,
        by_epoch=True,
        convert_to_iter_based=True),
]

# automatically scaling LR based on the actual training batch size
auto_scale_lr = dict(base_batch_size=1024)

# codec settings
codec = dict(
    type='SimCCLabel',
    input_size=(512, 512),
    sigma=(12, 12),
    simcc_split_ratio=2.0,
    normalize=False,
    use_dark=False)

# model settings
model = dict(
    type='TopdownPoseEstimator',
    data_preprocessor=dict(
        type='PoseDataPreprocessor',
        mean=[123.675, 116.28, 103.53],
        std=[58.395, 57.12, 57.375],
        bgr_to_rgb=True),
    backbone=dict(
        _scope_='mmdet',
        type='CSPNeXt',
        arch='P5',
        expand_ratio=0.5,
        deepen_factor=0.67,
        widen_factor=0.75,
        out_indices=(4, ),
        channel_attention=True,
        norm_cfg=dict(type='SyncBN'),
        act_cfg=dict(type='SiLU'),
        init_cfg=dict(
            type='Pretrained',
            prefix='backbone.',
            checkpoint='https://download.openmmlab.com/mmdetection/v3.0/rtmdet/cspnext_rsb_pretrain/cspnext-m_8xb256-rsb-a1-600e_in1k-ecb3bbd9.pth')),
    head=dict(
        type='RTMCCHead',
        in_channels=768,
        out_channels=38,
        input_size=codec['input_size'],
        in_featuremap_size=tuple([s // 32 for s in codec['input_size']]),
        simcc_split_ratio=codec['simcc_split_ratio'],
        final_layer_kernel_size=7,
        gau_cfg=dict(
            hidden_dims=256,
            s=128,
            expansion_factor=2,
            dropout_rate=0.,
            drop_path=0.,
            act_fn='SiLU',
            use_rel_bias=False,
            pos_enc=False),
        loss=dict(
            type='KLDiscretLoss',
            use_target_weight=True,
            beta=10.,
            label_softmax=True),
        decoder=codec),
    test_cfg=dict(flip_test=True))


# dataset settings
dataset_type = 'CephalometricDataset'
data_mode = 'topdown'
data_root = '/data/zhangHY/CL-Detection2023'


# meta_keys is used to add 'spacing' information, please do not change it if you don't know its usage
train_pipeline = [
    dict(type='LoadImage'),
    dict(type='GetBBoxCenterScale', padding=1.0),
    dict(type='RandomFlip', direction='horizontal'),
    dict(type='RandomBBoxTransform', scale_factor=[0.8, 1.2], rotate_factor=30),
    dict(type='TopdownAffine', input_size=codec['input_size']),
    dict(
        type='Albumentation',
        transforms=[
            dict(type='Blur', p=0.1),
            dict(type='MedianBlur', p=0.1),
            dict(
                type='CoarseDropout',
                max_holes=1,
                max_height=0.2,
                max_width=0.2,
                min_holes=1,
                min_height=0.1,
                min_width=0.1,
                p=1.),
        ]),
    dict(type='GenerateTarget', encoder=codec),
    dict(type='PackPoseInputs', meta_keys=('id', 'img_id', 'img_path', 'category_id', 'crowd_index', 'ori_shape',
                                           'img_shape', 'input_size', 'input_center', 'input_scale', 'flip',
                                           'flip_direction', 'flip_indices', 'raw_ann_info', 'spacing'))
]

val_pipeline = [
    dict(type='LoadImage'),
    dict(type='GetBBoxCenterScale'),
    dict(type='TopdownAffine', input_size=codec['input_size']),
    dict(type='PackPoseInputs', meta_keys=('id', 'img_id', 'img_path', 'category_id', 'crowd_index', 'ori_shape',
                                           'img_shape', 'input_size', 'input_center', 'input_scale', 'flip',
                                           'flip_direction', 'flip_indices', 'raw_ann_info', 'spacing'))
]

train_pipeline_stage2 = [
    dict(type='LoadImage', backend_args=backend_args),
    dict(type='GetBBoxCenterScale', padding=1.0),
    dict(type='RandomFlip', direction='horizontal'),
    dict(type='RandomBBoxTransform',
         shift_factor=0.,
         scale_factor=[0.9, 1.1],
         rotate_factor=10),
    dict(type='TopdownAffine', input_size=codec['input_size']),
    dict(
        type='Albumentation',
        transforms=[
            dict(type='Blur', p=0.1),
            dict(type='MedianBlur', p=0.1),
            dict(
                type='CoarseDropout',
                max_holes=1,
                max_height=0.2,
                max_width=0.2,
                min_holes=1,
                min_height=0.1,
                min_width=0.1,
                p=0.5),
        ]),
    dict(type='GenerateTarget', encoder=codec),
    dict(type='PackPoseInputs')
]

# dataloaders
train_dataloader = dict(
    batch_size=8,
    num_workers=2,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=True),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        data_mode='topdown',
        ann_file='train.json',
        data_prefix=dict(img='MMPose/'),
        pipeline=train_pipeline))

val_dataloader = dict(
    batch_size=4,
    num_workers=2,
    persistent_workers=True,
    drop_last=False,
    sampler=dict(type='DefaultSampler', shuffle=False, round_up=False),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        data_mode='topdown',
        ann_file='valid.json',
        data_prefix=dict(img='MMPose/'),
        test_mode=True,
        pipeline=val_pipeline))

test_dataloader = dict(
    batch_size=4,
    num_workers=2,
    persistent_workers=True,
    drop_last=False,
    sampler=dict(type='DefaultSampler', shuffle=False, round_up=False),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        data_mode='topdown',
        ann_file='test.json',
        data_prefix=dict(img='MMPose/'),
        test_mode=True,
        pipeline=val_pipeline))


# hooks
default_hooks = dict(
    timer=dict(type='IterTimerHook'),
    logger=dict(type='LoggerHook', interval=1),
    param_scheduler=dict(type='ParamSchedulerHook'),
    checkpoint=dict(
        type='CheckpointHook',
        interval=2,
        max_keep_ckpts=5,
        save_best='SDR 2.0mm',
        rule='greater'),
    sampler_seed=dict(type='DistSamplerSeedHook'),
    visualization=dict(type='PoseVisualizationHook', enable=False))

custom_hooks = [
    dict(
        type='EMAHook',
        ema_type='ExpMomentumEMA',
        momentum=0.0002,
        update_buffers=True,
        priority=49),
    dict(
        type='mmdet.PipelineSwitchHook',
        switch_epoch=max_epochs - stage2_num_epochs,
        switch_pipeline=train_pipeline_stage2)
]

val_evaluator = dict(
    type='CephalometricMetric')
test_evaluator = val_evaluator