wa1tzy的博客

train.py#dataset:https://www.crcv.ucf.edu/data/UCF101.phpimport timeitfrom datetime import datetimeimport socketimport osimport globfrom tqdm import tqdmimport torchfrom tensorboardX import SummaryWriterfrom torch import nn, optimfrom torch.ut

C3D网络架构与结构图：3D卷积与2D卷积的区别：其中多了一个时间维度三维卷积与三维池化的理解：① nn.Conv3d(3, 64, kernel_size=(3, 3, 3), padding=(1, 1, 1))kernel_size=(3, 3, 3)：第一个3：一共16帧，当前处理3帧；第二、三个3：H,W的大小；padding=(1, 1, 1)：三维、高、宽都填充1② nn.MaxPool3d(kernel_size=(1, 2, 2), stride=(1, 2, 2))

文章目录一、mypath.py二、datasets.py一、mypath.pyclass Path(object): @staticmethod def db_dir(database): if database == 'ucf101': # folder that contains class labels root_dir = r'F:\DATASET\UCF-101\UCF-101' # Sa