基于Alphapose生成的skeleton point,整理成stgcn需要的的格式（类似kinetics-skeleton），进入stgcn训练

stgcn的虚拟环境名字是torch
参考：https://blog.csdn.net/qq_43019451/article/details/118787580
注意：需要对pose数据归一化，在feeder_kinetics.py中121行有中心化的操作。
1、修改tools\kinetics_gendata.py中：

'--data_path', default='data/KTH/kinetics-skeleton'
'--out_folder', default='data/KTH/kinetics-skeleton'

def gendata(
        data_path,
        label_path,
        data_out_path,
        label_out_path,
        num_person_in=1,  #observe the first 5 persons
        num_person_out=1,  #then choose 2 persons with the highest score
        max_frame=300):

shape=(len(sample_name), 3, max_frame, 18, num_person_out))

2、修改feeder/feeder_kinetics.py中：

# sample_id = [name.split('.')[0] for name in self.sample_name]
sample_id = [name.split('.json')[0] for name in self.sample_name]
self.V = 26  # joint,根据不同的skeleton获得的节点数而定，coco是18个节点

3、python tools/kinetics_gendata.py, 生成:

train_data.npy
train_label.pkl
val_data.npy
val_label.pkl

4、在net/utils/graph.py文件里面get_edge函数中增加一个elif，num_node为 关键点个数、self_link为连接关系。如下添加的是一个‘my_pose’Layout，关键点个数为20（默认的pose点数是18）。

    elif layout == 'my_pose':
        self.num_node = 26
        self_link = [(i, i) for i in range(self.num_node)]
        neighbor_link = [(0, 1), (1, 3), (0, 2), (2, 4), (17, 18), (18, 6),
                         (18, 5), (6, 8), (8, 10), (5, 7),
                         (7, 9), (11, 19), (18, 19), (19, 12), (14, 12), (11, 13),
                         (14, 16), (16, 25), (25, 21), (25, 23), (13, 15), (15, 24),
                         (24, 20),(24,22)]
        self.edge = self_link + neighbor_link
        self.center = 18-19

5、修改config/st_gcn/kinetics-skeleton/train.yaml中的相关参数：

work_dir: ./work_dir/recognition/kinetics_skeleton/ST_GCN

# feeder
feeder: feeder.feeder.Feeder
train_feeder_args:
  random_choose: True
  random_move: True
  window_size: 150 # The length of the output sequence
  data_path: ./data/KTH/kinetics-skeleton/train_data.npy
  label_path: ./data/KTH/kinetics-skeleton/train_label.pkl
test_feeder_args:
  data_path: ./data/KTH/kinetics-skeleton/val_data.npy
  label_path: ./data/KTH/kinetics-skeleton/val_label.pkl

# model
model: net.st_gcn.Model
model_args:
  in_channels: 3
  num_class: 6
  edge_importance_weighting: True
  graph_args:
    layout: 'my_pose'
    strategy: 'spatial'

# training
device: [0]
batch_size: 32
test_batch_size: 32

#optim
base_lr: 0.1
step: [20, 30, 40, 50]
num_epoch: 500

6、执行训练：python main.py recognition -c config/st_gcn/kinetics-skeleton/train.yaml

7、在resource/kinetics_skeleton中，新建lable_name.txt,类别写入。 修改test.py中：

   def __init__(self, in_channels=3, num_class=6,
             edge_importance_weighting=True, **kwargs):
    super().__init__()
    # load graph
    self.graph = Graph(layout='my_pose', strategy='spatial',)


weights_path = 'work_dir/recognition/kinetics_skeleton/ST_GCN/epoch30_model.pt'
    # label_list = ['standing', 'walking', 'laying']
    label_list = []
    with open('resource/kinetics_skeleton/label_name_KTH.txt', 'r') as f:
        for line in f:
            label_list.append(line.strip('\n'))
    data_path = "data/KTH/kinetics-skeleton/val_data.npy"
    label_path = "data/KTH/kinetics-skeleton/val_label.pkl"

测试：python test.py,