【mmaction2 入门教程 04】训练 AVA 数据集中的自定义类别

0 前言

在时空行为数据集中，最常出现的就是长尾数据集，即某些类别的动作标签过少，导致训练效果不好，在mmation2当中，提供了一个方法，就是可以自定义要训练的类别。

那么先看看我之前训练的分析结果

上图中，深蓝色的样本数量，浅蓝色的ap值，可以看出，样本极少的类别，ap值几乎为0，那么我们在训练的时候，可以忽略这些行为。

那如何来做呢，在mmaction2文档中已经说明了：https://mmaction2.readthedocs.io/zh_CN/latest/detection_models.html#id11

GPU平台：https://cloud.videojj.com/auth/register?inviter=18452&activityChannel=student_invite

b站：https://www.bilibili.com/video/BV1RV4y1s7mK/

1（官网）训练 AVA 数据集中的自定义类别

用户可以训练 AVA 数据集中的自定义类别。AVA 中不同类别的样本量很不平衡：其中有超过 100000 样本的类别： stand/listen to (a person)/talk to (e.g., self, a person, a group)/watch (a person)，也有样本较少的类别（半数类别不足 500 样本）。大多数情况下，仅使用样本较少的类别进行训练将在这些类别上得到更好精度。

训练 AVA 数据集中的自定义类别包含 3 个步骤：

1. 从原先的类别中选择希望训练的类别，将其填写至配置文件的 custom_classes 域中。其中 0 不表示具体的动作类别，不应被选择。
2. 将 num_classes 设置为 num_classes = len(custom_classes) + 1。
   • 在新的类别到编号的对应中，编号 0 仍对应原类别 0，编号 i (i > 0) 对应原类别 custom_classes[i-1]。
   • 配置文件中 3 处涉及 num_classes 需要修改：model -> roi_head -> bbox_head -> num_classes， data -> train -> num_classes， data -> val -> num_classes.
   • 若 num_classes <= 5， 配置文件 BBoxHeadAVA 中的 topk 参数应被修改。topk 的默认值为 (3, 5)，topk 中的所有元素应小于 num_classes。
3. 确认所有自定义类别在 label_file 中。

2 训练数据集中的自定义类别

那么就开始咯，通过统计分析，原本71个动作类，选取了其中21个动作类。

2.1 配置文件

直接上配置文件：

# model setting
custom_classes = [3, 8, 11, 12, 13, 22, 23, 29, 32, 33, 36, 42, 45, 46, 49, 52, 55, 56, 59, 60, 67]
#custom_classes = [2, 7, 10, 11, 12, 21, 22, 28, 31, 32, 35, 41, 44, 45, 48, 51, 54, 55, 58, 59, 66]
num_classes = len(custom_classes) + 1
model = dict(
    type='FastRCNN',
    backbone=dict(
        type='ResNet3dSlowFast',
        pretrained=None,
        resample_rate=8,
        speed_ratio=8,
        channel_ratio=8,
        slow_pathway=dict(
            type='resnet3d',
            depth=50,
            pretrained=None,
            lateral=True,
            conv1_kernel=(1, 7, 7),
            dilations=(1, 1, 1, 1),
            conv1_stride_t=1,
            pool1_stride_t=1,
            inflate=(0, 0, 1, 1),
            spatial_strides=(1, 2, 2, 1)),
        fast_pathway=dict(
            type='resnet3d',
            depth=50,
            pretrained=None,
            lateral=False,
            base_channels=8,
            conv1_kernel=(5, 7, 7),
            conv1_stride_t=1,
            pool1_stride_t=1,
            spatial_strides=(1, 2, 2, 1))),
    roi_head=dict(
        type='AVARoIHead',
        bbox_roi_extractor=dict(
            type='SingleRoIExtractor3D',
            roi_layer_type='RoIAlign',
            output_size=8,
            with_temporal_pool=True),
        bbox_head=dict(
            type='BBoxHeadAVA',
            in_channels=2304,
            #num_classes=81,
            num_classes=num_classes,
            multilabel=True,
            dropout_ratio=0.5)),
    train_cfg=dict(
        rcnn=dict(
            assigner=dict(
                type='MaxIoUAssignerAVA',
                pos_iou_thr=0.9,
                neg_iou_thr=0.9,
                min_pos_iou=0.9),
            sampler=dict(
                type='RandomSampler',
                num=32,
                pos_fraction=1,
                neg_pos_ub=-1,
                add_gt_as_proposals=True),
            pos_weight=1.0,
            debug=False)),
    test_cfg=dict(rcnn=dict(action_thr=0.002)))

dataset_type = 'AVADataset'
data_root = '/home/MPCLST/Dataset/rawframes'
anno_root = '/home/MPCLST/Dataset/annotations'


#ann_file_train = f'{anno_root}/ava_train_v2.1.csv'
ann_file_train = f'{anno_root}/train.csv'
#ann_file_val = f'{anno_root}/ava_val_v2.1.csv'
ann_file_val = f'{anno_root}/val.csv'

#exclude_file_train = f'{anno_root}/ava_train_excluded_timestamps_v2.1.csv'
#exclude_file_val = f'{anno_root}/ava_val_excluded_timestamps_v2.1.csv'

exclude_file_train = f'{anno_root}/train_excluded_timestamps.csv'
exclude_file_val = f'{anno_root}/val_excluded_timestamps.csv'

#label_file = f'{anno_root}/ava_action_list_v2.1_for_activitynet_2018.pbtxt'
label_file = f'{anno_root}/action_list.pbtxt'

proposal_file_train = (f'{anno_root}/dense_proposals_train.pkl')
proposal_file_val = f'{anno_root}/dense_proposals_val.pkl'

img_norm_cfg = dict(
    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_bgr=False)

train_pipeline = [
    dict(type='SampleAVAFrames', clip_len=32, frame_interval=2),
    dict(type='RawFrameDecode'),
    dict(type='RandomRescale', scale_range=(256, 320)),
    dict(type='RandomCrop', size=256),
    dict(type='Flip', flip_ratio=0.5),
    dict(type='Normalize', **img_norm_cfg),
    dict(type='FormatShape', input_format='NCTHW', collapse=True),
    # Rename is needed to use mmdet detectors
    dict(type='Rename', mapping=dict(imgs='img')),
    dict(type='ToTensor', keys=['img', 'proposals', 'gt_bboxes', 'gt_labels']),
    dict(
        type='ToDataContainer',
        fields=[
            dict(key=['proposals', 'gt_bboxes', 'gt_labels'], stack=False)
        ]),
    dict(
        type='Collect',
        keys=['img', 'proposals', 'gt_bboxes', 'gt_labels'],
        meta_keys=['scores', 'entity_ids'])
]
# The testing is w/o. any cropping / flipping
val_pipeline = [
    dict(type='SampleAVAFrames', clip_len=32, frame_interval=2),
    dict(type='RawFrameDecode'),
    dict(type='Resize', scale=(-1, 256)),
    dict(type='Normalize', **img_norm_cfg),
    dict(type='FormatShape', input_format='NCTHW', collapse=True),
    # Rename is needed to use mmdet detectors
    dict(type='Rename', mapping=dict(imgs='img')),
    dict(type='ToTensor', keys=['img', 'proposals']),
    dict(type='ToDataContainer', fields=[dict(key='proposals', stack=False)]),
    dict(
        type='Collect',
        keys=['img', 'proposals'],
        meta_keys=['scores', 'img_shape'],
        nested=True)
]

data = dict(
    #videos_per_gpu=9,
    #workers_per_gpu=2,
    videos_per_gpu=5,
    workers_per_gpu=2,
    val_dataloader=dict(videos_per_gpu=1),
    test_dataloader=dict(videos_per_gpu=1),
    train=dict(
        type=dataset_type,
        ann_file=ann_file_train,
        exclude_file=exclude_file_train,
        pipeline=train_pipeline,
        label_file=label_file,
        proposal_file=proposal_file_train,
        person_det_score_thr=0.9,
        data_prefix=data_root,
        num_classes=num_classes,
        custom_classes=custom_classes,
        start_index=1,),
    val=dict(
        type=dataset_type,
        ann_file=ann_file_val,
        exclude_file=exclude_file_val,
        pipeline=val_pipeline,
        label_file=label_file,
        proposal_file=proposal_file_val,
        person_det_score_thr=0.9,
        data_prefix=data_root,
        num_classes=num_classes,
        custom_classes=custom_classes,
        start_index=1,))
data['test'] = data['val']

#optimizer = dict(type='SGD', lr=0.1125, momentum=0.9, weight_decay=0.00001)
optimizer = dict(type='SGD', lr=0.0125, momentum=0.9, weight_decay=0.00001)
# this lr is used for 8 gpus

optimizer_config = dict(grad_clip=dict(max_norm=40, norm_type=2))
# learning policy

lr_config = dict(
    policy='step',
    step=[10, 15],
    warmup='linear',
    warmup_by_epoch=True,
    warmup_iters=5,
    warmup_ratio=0.1)
#total_epochs = 20
total_epochs = 35
checkpoint_config = dict(interval=1)
workflow = [('train', 1)]
evaluation = dict(interval=1, save_best='mAP@0.5IOU')
log_config = dict(
    interval=20, hooks=[
        dict(type='TextLoggerHook'),
    ])
dist_params = dict(backend='nccl')
log_level = 'INFO'
work_dir = ('./work_dirs/22-8-15-custom-ava/'
            'slowfast_kinetics_pretrained_r50_4x16x1_20e_ava_rgb')
load_from = ('https://download.openmmlab.com/mmaction/recognition/slowfast/'
             'slowfast_r50_4x16x1_256e_kinetics400_rgb/'
             'slowfast_r50_4x16x1_256e_kinetics400_rgb_20200704-bcde7ed7.pth')
resume_from = None
find_unused_parameters = False

2.2 执行训练

#训练
cd /home/MPCLST/mmaction2_YF
python tools/train.py configs/detection/ava/my_custom_slowfast_kinetics_pretrained_r50_4x16x1_20e_ava.py --validate

2.3 可视化测试

在上一篇博客写过：【mmaction2 入门教程 03】评价指标可视化 mAP、每类行为的ap值、每类行为的数量

# 测试集的分析-latest
cd /home/MPCLST/mmaction2_YF/   
python tools/test.py configs/detection/ava/my_custom_slowfast_kinetics_pretrained_r50_4x16x1_20e_ava.py ./work_dirs/22-8-15-custom-ava/slowfast_kinetics_pretrained_r50_4x16x1_20e_ava_rgb/latest.pth --eval mAP

# 训练集的分析-latest
cd /home/MPCLST/mmaction2_YF/   
python tools/test.py configs/detection/ava/my_custom_slowfast_kinetics_pretrained_r50_4x16x1_20e_ava2.py ./work_dirs/22-8-15-custom-ava/slowfast_kinetics_pretrained_r50_4x16x1_20e_ava_rgb/latest.pth --eval mAP