name
This commit is contained in:
@ -1284,8 +1284,9 @@ def get_label(file_name, dataset_name):
|
||||
key = file_name[-5:]
|
||||
return HumanAct12[key]
|
||||
elif dataset_name == 'UTD_MHAD':
|
||||
key = file_name.split('_')[0][1:]
|
||||
return UTD_MHAD[key]
|
||||
###key = file_name.split('_')[0][1:]
|
||||
###return UTD_MHAD[key]
|
||||
return "single_person"
|
||||
elif dataset_name == 'CMU_Mocap':
|
||||
key = file_name.split(':')[0]
|
||||
return CMU_Mocap[key] if key in CMU_Mocap.keys() else ""
|
||||
|
||||
@ -1,25 +1,46 @@
|
||||
import scipy.io
|
||||
import numpy as np
|
||||
import json
|
||||
import json # 引入json模块
|
||||
|
||||
|
||||
def load(name, path):
|
||||
# 处理UTD-MHAD的JSON文件（你的单帧数据）
|
||||
if name == 'UTD_MHAD':
|
||||
arr = scipy.io.loadmat(path)['d_skel']
|
||||
new_arr = np.zeros([arr.shape[2], arr.shape[0], arr.shape[1]])
|
||||
for i in range(arr.shape[2]):
|
||||
for j in range(arr.shape[0]):
|
||||
for k in range(arr.shape[1]):
|
||||
new_arr[i][j][k] = arr[j][k][i]
|
||||
return new_arr
|
||||
# 判断文件是否为JSON格式（通过后缀）
|
||||
if path.endswith('.json'):
|
||||
with open(path, 'r') as f:
|
||||
data = json.load(f) # 加载JSON列表
|
||||
# 转换为NumPy数组，确保形状为[关节数, 3]
|
||||
data_np = np.array(data)
|
||||
# 校验数据格式（防止错误）
|
||||
assert data_np.ndim == 2 and data_np.shape[1] == 3, \
|
||||
f"UTD-MHAD JSON格式错误，应为[关节数, 3]，实际为{data_np.shape}"
|
||||
# 若需要单帧维度（[1, 关节数, 3]），可扩展维度
|
||||
return data_np[np.newaxis, ...] # 输出形状：[1, N, 3]（1表示单帧）
|
||||
|
||||
# 保留原UTD_MHAD的.mat文件支持（如果还需要处理.mat数据）
|
||||
elif path.endswith('.mat'):
|
||||
arr = scipy.io.loadmat(path)['d_skel']
|
||||
new_arr = np.zeros([arr.shape[2], arr.shape[0], arr.shape[1]])
|
||||
for i in range(arr.shape[2]):
|
||||
for j in range(arr.shape[0]):
|
||||
for k in range(arr.shape[1]):
|
||||
new_arr[i][j][k] = arr[j][k][i]
|
||||
return new_arr
|
||||
|
||||
else:
|
||||
raise ValueError(f"UTD-MHAD不支持的文件格式：{path}")
|
||||
|
||||
# 其他数据集的原有逻辑保持不变
|
||||
elif name == 'HumanAct12':
|
||||
return np.load(path, allow_pickle=True)
|
||||
elif name == "CMU_Mocap":
|
||||
return np.load(path, allow_pickle=True)
|
||||
elif name == "Human3.6M":
|
||||
return np.load(path, allow_pickle=True)[0::5] # down_sample
|
||||
return np.load(path, allow_pickle=True)[0::5] # 下采样
|
||||
elif name == "NTU":
|
||||
return np.load(path, allow_pickle=True)[0::2]
|
||||
elif name == "HAA4D":
|
||||
return np.load(path, allow_pickle=True)
|
||||
|
||||
else:
|
||||
raise ValueError(f"不支持的数据集名称：{name}")
|
||||
@ -1,53 +1,104 @@
|
||||
"""
|
||||
SMPL模型拟合主程序
|
||||
该脚本用于将人体姿态数据拟合到SMPL（Skinned Multi-Person Linear）模型中
|
||||
主要功能：
|
||||
1. 加载人体姿态数据
|
||||
2. 使用SMPL模型进行拟合优化
|
||||
3. 保存拟合结果和可视化图像
|
||||
"""
|
||||
|
||||
import os
|
||||
import sys
|
||||
sys.path.append(os.getcwd())
|
||||
from meters import Meters
|
||||
from smplpytorch.pytorch.smpl_layer import SMPL_Layer
|
||||
from train import train
|
||||
from transform import transform
|
||||
from save import save_pic, save_params
|
||||
from load import load
|
||||
import torch
|
||||
import numpy as np
|
||||
from tensorboardX import SummaryWriter
|
||||
from easydict import EasyDict as edict
|
||||
import time
|
||||
import logging
|
||||
|
||||
import argparse
|
||||
import json
|
||||
# 导入自定义模块
|
||||
from meters import Meters # 用于跟踪训练指标的工具类
|
||||
from smplpytorch.pytorch.smpl_layer import SMPL_Layer # SMPL模型层
|
||||
from train import train # 训练函数
|
||||
from transform import transform # 数据变换函数
|
||||
from save import save_pic, save_params # 保存结果的函数
|
||||
from load import load # 数据加载函数
|
||||
|
||||
# 导入标准库
|
||||
import torch # PyTorch深度学习框架
|
||||
import numpy as np # 数值计算库
|
||||
from easydict import EasyDict as edict # 用于创建字典对象的便捷工具
|
||||
import time # 时间处理
|
||||
import logging # 日志记录
|
||||
|
||||
import argparse # 命令行参数解析
|
||||
import json # JSON文件处理
|
||||
|
||||
# 启用CUDNN加速，提高训练效率
|
||||
torch.backends.cudnn.enabled = True
|
||||
torch.backends.cudnn.benchmark = True
|
||||
|
||||
|
||||
def parse_args():
|
||||
"""
|
||||
解析命令行参数
|
||||
|
||||
Returns:
|
||||
args: 包含解析后参数的命名空间对象
|
||||
"""
|
||||
parser = argparse.ArgumentParser(description='Fit SMPL')
|
||||
|
||||
# 实验名称，默认使用当前时间戳
|
||||
parser.add_argument('--exp', dest='exp',
|
||||
help='Define exp name',
|
||||
default=time.strftime('%Y-%m-%d %H-%M-%S', time.localtime(time.time())), type=str)
|
||||
|
||||
# 数据集名称，用于选择对应的配置文件
|
||||
parser.add_argument('--dataset_name', '-n', dest='dataset_name',
|
||||
help='select dataset',
|
||||
default='', type=str)
|
||||
|
||||
# 数据集路径，可以覆盖配置文件中的默认路径
|
||||
parser.add_argument('--dataset_path', dest='dataset_path',
|
||||
help='path of dataset',
|
||||
default=None, type=str)
|
||||
|
||||
args = parser.parse_args()
|
||||
return args
|
||||
|
||||
|
||||
def get_config(args):
|
||||
"""
|
||||
根据数据集名称加载对应的配置文件
|
||||
|
||||
Args:
|
||||
args: 命令行参数对象
|
||||
|
||||
Returns:
|
||||
cfg: 配置对象，包含所有训练和模型参数
|
||||
"""
|
||||
# 根据数据集名称构建配置文件路径
|
||||
config_path = 'fit/configs/{}.json'.format(args.dataset_name)
|
||||
|
||||
# 读取JSON配置文件
|
||||
with open(config_path, 'r') as f:
|
||||
data = json.load(f)
|
||||
|
||||
# 将字典转换为edict对象，支持点号访问属性
|
||||
cfg = edict(data.copy())
|
||||
|
||||
# 如果命令行指定了数据集路径，则覆盖配置文件中的设置
|
||||
if not args.dataset_path == None:
|
||||
cfg.DATASET.PATH = args.dataset_path
|
||||
|
||||
return cfg
|
||||
|
||||
|
||||
def set_device(USE_GPU):
|
||||
"""
|
||||
根据配置和硬件可用性设置计算设备
|
||||
|
||||
Args:
|
||||
USE_GPU: 是否使用GPU的布尔值
|
||||
|
||||
Returns:
|
||||
device: PyTorch设备对象（'cuda' 或 'cpu'）
|
||||
"""
|
||||
if USE_GPU and torch.cuda.is_available():
|
||||
device = torch.device('cuda')
|
||||
else:
|
||||
@ -56,9 +107,21 @@ def set_device(USE_GPU):
|
||||
|
||||
|
||||
def get_logger(cur_path):
|
||||
"""
|
||||
设置日志记录器，同时输出到文件和控制台
|
||||
|
||||
Args:
|
||||
cur_path: 当前实验路径，用于保存日志文件
|
||||
|
||||
Returns:
|
||||
logger: 日志记录器对象
|
||||
writer: TensorBoard写入器（当前设置为None）
|
||||
"""
|
||||
# 创建日志记录器
|
||||
logger = logging.getLogger(__name__)
|
||||
logger.setLevel(level=logging.INFO)
|
||||
|
||||
# 设置文件输出处理器，将日志保存到log.txt文件
|
||||
handler = logging.FileHandler(os.path.join(cur_path, "log.txt"))
|
||||
handler.setLevel(logging.INFO)
|
||||
formatter = logging.Formatter(
|
||||
@ -66,6 +129,7 @@ def get_logger(cur_path):
|
||||
handler.setFormatter(formatter)
|
||||
logger.addHandler(handler)
|
||||
|
||||
# 设置控制台输出处理器，将日志同时输出到终端
|
||||
handler = logging.StreamHandler()
|
||||
handler.setLevel(logging.INFO)
|
||||
formatter = logging.Formatter(
|
||||
@ -73,54 +137,106 @@ def get_logger(cur_path):
|
||||
handler.setFormatter(formatter)
|
||||
logger.addHandler(handler)
|
||||
|
||||
writer = SummaryWriter(os.path.join(cur_path, 'tb'))
|
||||
# TensorBoard写入器（目前被注释掉，设置为None）
|
||||
# from tensorboardX import SummaryWriter
|
||||
# writer = SummaryWriter(os.path.join(cur_path, 'tb'))
|
||||
writer = None
|
||||
|
||||
return logger, writer
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
"""
|
||||
主函数：执行SMPL模型拟合流程
|
||||
|
||||
主要步骤：
|
||||
1. 解析命令行参数
|
||||
2. 创建实验目录
|
||||
3. 加载配置文件
|
||||
4. 设置日志记录
|
||||
5. 初始化SMPL模型
|
||||
6. 遍历数据集进行拟合
|
||||
7. 保存结果
|
||||
"""
|
||||
# 解析命令行参数
|
||||
args = parse_args()
|
||||
|
||||
# 创建实验目录，使用时间戳或用户指定的实验名称
|
||||
cur_path = os.path.join(os.getcwd(), 'exp', args.exp)
|
||||
assert not os.path.exists(cur_path), 'Duplicate exp name'
|
||||
assert not os.path.exists(cur_path), 'Duplicate exp name' # 确保实验名称不重复
|
||||
os.mkdir(cur_path)
|
||||
|
||||
# 加载配置文件
|
||||
cfg = get_config(args)
|
||||
# 将配置保存到实验目录中，便于后续追踪实验设置
|
||||
json.dump(dict(cfg), open(os.path.join(cur_path, 'config.json'), 'w'))
|
||||
|
||||
# 设置日志记录器
|
||||
logger, writer = get_logger(cur_path)
|
||||
logger.info("Start print log")
|
||||
|
||||
# 设置计算设备（GPU或CPU）
|
||||
device = set_device(USE_GPU=cfg.USE_GPU)
|
||||
logger.info('using device: {}'.format(device))
|
||||
|
||||
# 初始化SMPL模型层
|
||||
# center_idx=0: 设置中心关节点索引
|
||||
# gender='male': 设置性别为男性（注释掉的cfg.MODEL.GENDER可能用于从配置文件读取）
|
||||
# model_root: SMPL模型文件的路径
|
||||
smpl_layer = SMPL_Layer(
|
||||
center_idx=0,
|
||||
gender=cfg.MODEL.GENDER,
|
||||
gender='male', #cfg.MODEL.GENDER,
|
||||
model_root='smplpytorch/native/models')
|
||||
|
||||
# 初始化指标记录器，用于跟踪训练损失等指标
|
||||
meters = Meters()
|
||||
file_num = 0
|
||||
# 遍历数据集目录中的所有文件
|
||||
for root, dirs, files in os.walk(cfg.DATASET.PATH):
|
||||
for file in sorted(files):
|
||||
if not 'baseball_swing' in file:
|
||||
continue
|
||||
file_num += 1
|
||||
for file in sorted(files): # 按文件名排序处理
|
||||
# 可选的文件过滤器（当前被注释掉）
|
||||
# 可以用于只处理特定的文件，如包含'baseball_swing'的文件
|
||||
###if not 'baseball_swing' in file:
|
||||
###continue
|
||||
|
||||
file_num += 1 # 文件计数器
|
||||
logger.info(
|
||||
'Processing file: {} [{} / {}]'.format(file, file_num, len(files)))
|
||||
|
||||
# 加载并变换目标数据
|
||||
# 1. load(): 根据数据集类型加载原始数据
|
||||
# 2. transform(): 将数据转换为模型所需的格式
|
||||
# 3. torch.from_numpy(): 将numpy数组转换为PyTorch张量
|
||||
# 4. .float(): 确保数据类型为float32
|
||||
target = torch.from_numpy(transform(args.dataset_name, load(args.dataset_name,
|
||||
os.path.join(root, file)))).float()
|
||||
logger.info("target shape:{}".format(target.shape))
|
||||
|
||||
# 执行SMPL模型拟合训练
|
||||
# 传入SMPL层、目标数据、日志记录器、设备信息等
|
||||
res = train(smpl_layer, target,
|
||||
logger, writer, device,
|
||||
args, cfg, meters)
|
||||
|
||||
# 更新平均损失指标
|
||||
# k=target.shape[0] 表示批次大小，用于加权平均
|
||||
meters.update_avg(meters.min_loss, k=target.shape[0])
|
||||
|
||||
# 重置早停计数器，为下一个文件的训练做准备
|
||||
meters.reset_early_stop()
|
||||
|
||||
# 记录当前的平均损失
|
||||
logger.info("avg_loss:{:.4f}".format(meters.avg))
|
||||
|
||||
# 保存拟合结果
|
||||
# 1. save_params(): 保存拟合得到的SMPL参数
|
||||
# 2. save_pic(): 保存可视化图像，包括拟合结果和原始目标的对比
|
||||
save_params(res, file, logger, args.dataset_name)
|
||||
save_pic(res, smpl_layer, file, logger, args.dataset_name, target)
|
||||
|
||||
# 清空GPU缓存，防止内存溢出
|
||||
torch.cuda.empty_cache()
|
||||
|
||||
# 所有文件处理完成，记录最终的平均损失
|
||||
logger.info(
|
||||
"Fitting finished! Average loss: {:.9f}".format(meters.avg))
|
||||
|
||||
@ -78,9 +78,9 @@ def train(smpl_layer, target,
|
||||
# epoch, float(loss),float(scale)))
|
||||
print("Epoch {}, lossPerBatch={:.6f}, scale={:.4f}".format(
|
||||
epoch, float(loss),float(scale)))
|
||||
writer.add_scalar('loss', float(loss), epoch)
|
||||
writer.add_scalar('learning_rate', float(
|
||||
optimizer.state_dict()['param_groups'][0]['lr']), epoch)
|
||||
###writer.add_scalar('loss', float(loss), epoch)
|
||||
###writer.add_scalar('learning_rate', float(
|
||||
###optimizer.state_dict()['param_groups'][0]['lr']), epoch)
|
||||
# save_single_pic(res,smpl_layer,epoch,logger,args.dataset_name,target)
|
||||
|
||||
logger.info('Train ended, min_loss = {:.4f}'.format(
|
||||
|
||||
@ -3,7 +3,7 @@ import numpy as np
|
||||
rotate = {
|
||||
'HumanAct12': [1., -1., -1.],
|
||||
'CMU_Mocap': [0.05, 0.05, 0.05],
|
||||
'UTD_MHAD': [-1., 1., -1.],
|
||||
'UTD_MHAD': [1., 1., 1.],
|
||||
'Human3.6M': [-0.001, -0.001, 0.001],
|
||||
'NTU': [1., 1., -1.],
|
||||
'HAA4D': [1., -1., -1.],
|
||||
|
||||
Reference in New Issue
Block a user