windows单GPU训练pyskl参数修改

configs文件posec3d/c3d_light_gym/joint

model = dict(
    type='Recognizer3D',
    backbone=dict(
        type='C3D',
        # 更改输入关键点数量13，则in_channels=13
        in_channels=17,
        base_channels=32,
        num_stages=3,
        temporal_downsample=False),
    cls_head=dict(
        type='I3DHead',
        in_channels=256,
        # 检测的行为类别，假如为4类，则num_classes=4
        num_classes=99,
        dropout=0.5),
    test_cfg=dict(average_clips='prob'))

dataset_type = 'PoseDataset'
# 训练数据的文件位置，pkl文件包含了训练数据和测试数据，数据集制作参考mmaction2 1.2.0
ann_file = 'D:/pycharm/pyskl/gym_hrnet.pkl' 
# 17个关键点中左关键点，0默认中位，去除脸部关键点可删除13，15
left_kp = [1, 3, 5, 7, 9, 11, 13, 15]
# 17个关键点中右关键点，0默认中位，去除脸部关键点可删除14.16
right_kp = [2, 4, 6, 8, 10, 12, 14, 16]
train_pipeline = [
    # 时间序列位48帧
    dict(type='UniformSampleFrames', clip_len=48),
    dict(type='PoseDecode'),
    dict(type='PoseCompact', hw_ratio=1., allow_imgpad=True),
    # 训练的热图映射到64x64大小
    dict(type='Resize', scale=(-1, 64)),
    # 数据增强
    dict(type='RandomResizedCrop', area_range=(0.56, 1.0)),
    dict(type='Resize', scale=(56, 56), keep_ratio=False),
    dict(type='Flip', flip_ratio=0.5, left_kp=left_kp, right_kp=right_kp),
    dict(type='GeneratePoseTarget', with_kp=True, with_limb=False),
    dict(type='FormatShape', input_format='NCTHW_Heatmap'),
    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
    dict(type='ToTensor', keys=['imgs', 'label'])
]
val_pipeline = [
    dict(type='UniformSampleFrames', clip_len=48, num_clips=1, test_mode=True),
    dict(type='PoseDecode'),
    dict(type='PoseCompact', hw_ratio=1., allow_imgpad=True),
    dict(type='Resize', scale=(64, 64), keep_ratio=False),
    dict(type='GeneratePoseTarget', with_kp=True, with_limb=False),
    dict(type='FormatShape', input_format='NCTHW_Heatmap'),
    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
    dict(type='ToTensor', keys=['imgs'])
]
test_pipeline = [
    dict(type='UniformSampleFrames', clip_len=48, num_clips=10, test_mode=True),
    dict(type='PoseDecode'),
    dict(type='PoseCompact', hw_ratio=1., allow_imgpad=True),
    dict(type='Resize', scale=(64, 64), keep_ratio=False),
    dict(type='GeneratePoseTarget', with_kp=True, with_limb=False, double=True, left_kp=left_kp, right_kp=right_kp),
    dict(type='FormatShape', input_format='NCTHW_Heatmap'),
    dict(type='Collect', keys=['imgs', 'label'], meta_keys=[]),
    dict(type='ToTensor', keys=['imgs'])
]
data = dict(
    # 可以只看做是batch_size，分布式训练的单GPUbatch
    videos_per_gpu=1,
    # 可以只看做是num_workers   
    workers_per_gpu=8,
    test_dataloader=dict(videos_per_gpu=1),
    # split='train'和split='val'需要根据实际分配名称修改
    train=dict(
        type='RepeatDataset',
        times=10,
        dataset=dict(type=dataset_type, ann_file=ann_file, split='train', pipeline=train_pipeline)),
    val=dict(type=dataset_type, ann_file=ann_file, split='val', pipeline=val_pipeline),
    test=dict(type=dataset_type, ann_file=ann_file, split='val', pipeline=test_pipeline))
# optimizer
optimizer = dict(type='SGD', lr=0.4, momentum=0.9, weight_decay=0.0003)
optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy
lr_config = dict(policy='CosineAnnealing', by_epoch=False, min_lr=0)
# epochs
total_epochs = 24
# 保存权重间隔，建议interval设置一致，先设置1查看需要迭代数，再设置即可显示每轮结果
checkpoint_config = dict(interval=1)
evaluation = dict(interval=1, metrics=['top_k_accuracy', 'mean_class_accuracy'], topk=(1, 5))
# 日志打印
log_config = dict(interval=20, hooks=[dict(type='TextLoggerHook')])
log_level = 'INFO'
# 训练结果的保存位置
work_dir = './work_dirs/posec3d/c3d_light_gym/joint'