读源码：MMaction2-slowfast

最新推荐文章于 2024-01-29 17:39:40 发布
最新推荐文章于 2024-01-29 17:39:40 发布
阅读量574 收藏 11
点赞数 2
分类专栏： Pytorch源码学习
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/bairw_Bella/article/details/114850684
版权

mmaction2_SlowFast源码解读

ResNet3dSlowFast

class ResNet3dSlowFast(nn.Module):

    def __init__(self,
                 pretrained,
                 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)),
                 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)):
        super().__init__()

参数注解：

Args:
 depth (int): Depth of resnet, from {18, 34, 50, 101, 152}.
 pretrained (str): 预训练模型的文件路径
 resample_rate (int): 重采样的时序步长，指每`resample_rate`帧，处理一帧，默认值为16
 speed_ratio (int): 快慢路径之间的时间比率。Default: 8
 channel_ratio (int): 快路径通道数的降维 by ``channel_ratio``,  Default: 8.
 slow_pathway (dict): 慢路径的必要参数配置

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)),
type (str): the pathway 以哪个骨干网络为基础.
lateral (bool): 是否构建路径之间的潜在连接。

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)):

1，邻近差值改变输入张量的时序维度

iNPUT [4,3,32,224,224]
插值后 [4,3,8,224,224]

x_slow = nn.functional.interpolate(x,
            mode='nearest',
            scale_factor=(1.0 / self.resample_rate, 1.0, 1.0))

nn.functional.interpolate(x, mode='nearest', scale_factor=(1.0 / self.resample_rate, 1.0, 1.0)
对应pytorch官方文档函数：
torch.nn.functional.interpolate(input, size=None, scale_factor=None, mode='nearest', align_corners=None, recompute_scale_factor=None)

  1. 根据给定的尺寸或尺度因子进行上采样或下采样
  2. mode (str) –使用的上采样算法有: ‘nearest’ | ‘linear’ | ‘bilinear’ | ‘bicubic’ | ‘trilinear’ | ‘area’. 默认: ‘nearest’
    其中卷积层的初始化方式是Kaiming Initialization
    Xavier Initialization和Kaiming Initialization的motivation是使信号强度（使用Variance度量）在神经网络训练的过程中保持不变。

2, SlowPath

x_slow = self.slow_path.conv1(x_slow)
x_slow = self.slow_path.maxpool(x_slow)
  • 慢路由定义：
    forward：self.slow_path.conv1
    →
    slowpath构建
#slowpathway字典参数定义
 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)),
#根据slow_pathway参数构建路径
self.slow_path = build_pathway(slow_pathway)

def build_pathway(cfg, *args, **kwargs):
    """Build pathway.
    Args:传入路径配置参数（字典类型）
        cfg (None or dict): cfg should contain:
            - type (str): identify conv layer type.
    Returns: 返回卷积模块构建卷积路由
        nn.Module: Created pathway.
    """
    #cfg配置参数应是字典类型
    if not (isinstance(cfg, dict) and 'type' in cfg):
        raise TypeError('cfg must be a dict containing the key "type"')
    cfg_ = cfg.copy()
"""
预先定义pathway可选的骨干网络类型，mmaction2目前只实现了ResNet骨干网络
pathway_cfg = {
    'resnet3d': ResNet3dPathway,
    # TODO: BNInceptionPathway
}
"""
    pathway_type = cfg_.pop('type')
    if pathway_type not in pathway_cfg:
        raise KeyError(f'Unrecognized pathway type {pathway_type}')
    else:
        pathway_cls = pathway_cfg[pathway_type]
    #获取ResNet3dPathway模块
    pathway = pathway_cls(*args, **kwargs, **cfg_)

    return pathway

→获取SlowPath模块

  • ResNet3dPathway 参数
class ResNet3dPathway(ResNet3d):
    """A pathway of Slowfast based on ResNet3d.

    Args:
        *args (arguments): 参数同 :class:``ResNet3d``.
        lateral (bool): 确定是否使用两个路由之间的潜在连接。默认：False
        speed_ratio (int): Speed ratio indicating the ratio between time
            dimension of the fast and slow pathway,Default: 8.
        channel_ratio (int): 快路由的降维。Default: 8.
        fusion_kernel (int):潜在融合的核尺寸.Default: 5.
        **kwargs (keyword arguments): Keywork arguments for ResNet3d.
    """
  • Slow模块
    Conv1：self.slow_path.conv1
'''
ConvModule(
  (conv): Conv3d(3, 64, kernel_size=(1, 7, 7), stride=(1, 2, 2), padding=(0, 3, 3), bias=False)
  (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (activate): ReLU(inplace=True)
)
'''

#融合快路由的特征到慢路由中，卷积的输入输出通道设置符合快慢路由之间的通道比率
self.conv1_lateral = ConvModule(
                self.inplanes // self.channel_ratio,
                # https://arxiv.org/abs/1812.03982, the
                # third type of lateral connection has out_channel:
                # 2 * \beta * C
                self.inplanes * 2 // self.channel_ratio,
                kernel_size=(fusion_kernel, 1, 1),
                stride=(self.speed_ratio, 1, 1),
                padding=((fusion_kernel - 1) // 2, 0, 0),
                bias=False,
                conv_cfg=self.conv_cfg,
                norm_cfg=None,
                act_cfg=None)

'''
#Conv1_lateral
ConvModule(
  (conv): Conv3d(8, 16, kernel_size=(5, 1, 1), stride=(8, 1, 1), padding=(2, 0, 0), bias=False)
) 
'''

self.slow_path.maxpool

MaxPool3d(kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), dilation=1, ceil_mode=False)
  • Fastpath模块
  (fast_path): ResNet3dPathway(
    (conv1): ConvModule(
      (conv): Conv3d(3, 8, kernel_size=(5, 7, 7), stride=(1, 2, 2), padding=(2, 3, 3), bias=False)
      (bn): BatchNorm3d(8, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (activate): ReLU(inplace=True)
    )
    (maxpool): MaxPool3d(kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), dilation=1, ceil_mode=False)
    (pool2): MaxPool3d(kernel_size=(2, 1, 1), stride=(2, 1, 1), padding=0, dilation=1, ceil_mode=False)

残差模块层 Layer1_Lateral

ConvModule(
  (conv): Conv3d(32, 64, kernel_size=(5, 1, 1), stride=(8, 1, 1), padding=(2, 0, 0), bias=False)
)
  • 整个SlowFast的网络结构
ResNet3dSlowFast(
  (slow_path): ResNet3dPathway(
    (conv1): ConvModule(
      (conv): Conv3d(3, 64, kernel_size=(1, 7, 7), stride=(1, 2, 2), padding=(0, 3, 3), bias=False)
      (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (activate): ReLU(inplace=True)
    )
    (maxpool): MaxPool3d(kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), dilation=1, ceil_mode=False)
    (pool2): MaxPool3d(kernel_size=(2, 1, 1), stride=(2, 1, 1), padding=0, dilation=1, ceil_mode=False)
    (layer1): Sequential(
      (0): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(80, 64, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(64, 64, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(64, 256, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (downsample): ConvModule(
          (conv): Conv3d(80, 256, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (1): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(256, 64, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(64, 64, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(64, 256, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (2): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(256, 64, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(64, 64, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(64, 256, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
    )
    (layer2): Sequential(
      (0): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(320, 128, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(128, 128, kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(128, 512, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (downsample): ConvModule(
          (conv): Conv3d(320, 512, kernel_size=(1, 1, 1), stride=(1, 2, 2), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (1): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(512, 128, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(128, 128, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(128, 512, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (2): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(512, 128, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(128, 128, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(128, 512, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (3): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(512, 128, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(128, 128, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(128, 512, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
    )
    (layer3): Sequential(
      (0): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(640, 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(256, 256, kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(256, 1024, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (downsample): ConvModule(
          (conv): Conv3d(640, 1024, kernel_size=(1, 1, 1), stride=(1, 2, 2), bias=False)
          (bn): BatchNorm3d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (1): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(1024, 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(256, 256, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(256, 1024, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (2): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(1024, 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(256, 256, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(256, 1024, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (3): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(1024, 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(256, 256, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(256, 1024, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (4): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(1024, 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(256, 256, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(256, 1024, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (5): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(1024, 256, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(256, 256, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(256, 1024, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
    )
    (layer4): Sequential(
      (0): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(1280, 512, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(512, 512, kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(512, 2048, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (downsample): ConvModule(
          (conv): Conv3d(1280, 2048, kernel_size=(1, 1, 1), stride=(1, 2, 2), bias=False)
          (bn): BatchNorm3d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (1): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(2048, 512, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(512, 512, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(512, 2048, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (2): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(2048, 512, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(512, 512, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(512, 2048, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
    )
    (conv1_lateral): ConvModule(
      (conv): Conv3d(8, 16, kernel_size=(5, 1, 1), stride=(8, 1, 1), padding=(2, 0, 0), bias=False)
    )
    (layer1_lateral): ConvModule(
      (conv): Conv3d(32, 64, kernel_size=(5, 1, 1), stride=(8, 1, 1), padding=(2, 0, 0), bias=False)
    )
    (layer2_lateral): ConvModule(
      (conv): Conv3d(64, 128, kernel_size=(5, 1, 1), stride=(8, 1, 1), padding=(2, 0, 0), bias=False)
    )
    (layer3_lateral): ConvModule(
      (conv): Conv3d(128, 256, kernel_size=(5, 1, 1), stride=(8, 1, 1), padding=(2, 0, 0), bias=False)
    )
  )
  (fast_path): ResNet3dPathway(
    (conv1): ConvModule(
      (conv): Conv3d(3, 8, kernel_size=(5, 7, 7), stride=(1, 2, 2), padding=(2, 3, 3), bias=False)
      (bn): BatchNorm3d(8, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (activate): ReLU(inplace=True)
    )
    (maxpool): MaxPool3d(kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), dilation=1, ceil_mode=False)
    (pool2): MaxPool3d(kernel_size=(2, 1, 1), stride=(2, 1, 1), padding=0, dilation=1, ceil_mode=False)
    (layer1): Sequential(
      (0): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(8, 8, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(8, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(8, 8, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(8, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(8, 32, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (downsample): ConvModule(
          (conv): Conv3d(8, 32, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (1): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(32, 8, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(8, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(8, 8, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(8, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(8, 32, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (2): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(32, 8, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(8, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(8, 8, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(8, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(8, 32, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
    )
    (layer2): Sequential(
      (0): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(32, 16, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(16, 16, kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(16, 64, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (downsample): ConvModule(
          (conv): Conv3d(32, 64, kernel_size=(1, 1, 1), stride=(1, 2, 2), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (1): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(64, 16, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(16, 16, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(16, 64, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (2): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(64, 16, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(16, 16, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(16, 64, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (3): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(64, 16, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(16, 16, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(16, 64, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
    )
    (layer3): Sequential(
      (0): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(64, 32, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(32, 32, kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(32, 128, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (downsample): ConvModule(
          (conv): Conv3d(64, 128, kernel_size=(1, 1, 1), stride=(1, 2, 2), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (1): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(128, 32, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(32, 32, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(32, 128, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (2): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(128, 32, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(32, 32, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(32, 128, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (3): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(128, 32, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(32, 32, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(32, 128, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (4): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(128, 32, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(32, 32, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(32, 128, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (5): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(128, 32, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(32, 32, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(32, 128, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
    )
    (layer4): Sequential(
      (0): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(128, 64, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(64, 64, kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(64, 256, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (downsample): ConvModule(
          (conv): Conv3d(128, 256, kernel_size=(1, 1, 1), stride=(1, 2, 2), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (1): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(256, 64, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(64, 64, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(64, 256, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
      (2): Bottleneck3d(
        (conv1): ConvModule(
          (conv): Conv3d(256, 64, kernel_size=(3, 1, 1), stride=(1, 1, 1), padding=(1, 0, 0), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv2): ConvModule(
          (conv): Conv3d(64, 64, kernel_size=(1, 3, 3), stride=(1, 1, 1), padding=(0, 1, 1), bias=False)
          (bn): BatchNorm3d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
          (activate): ReLU(inplace=True)
        )
        (conv3): ConvModule(
          (conv): Conv3d(64, 256, kernel_size=(1, 1, 1), stride=(1, 1, 1), bias=False)
          (bn): BatchNorm3d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (relu): ReLU(inplace=True)
      )
    )
  )
)
Bella_wanna_Better
关注
  • 2
    点赞
  • 踩
  • 11
    收藏
    觉得还不错? 一键收藏
  • 1
    评论
专栏目录
22-8-6 mmaction2 slowfast训练配置 训练日志分析
08-06
包含配置文件： 训练日志：20220805_165139.log.json 训练配置：my_slowfast_kinetics_pretrained_r50_4x16x1_20e_ava_rgb.py 训练配置（为了测试训练集的效果）：my_slowfast_kinetics_pretrained_r50_4x16x1_20e_ava_rgb2.py
动作识别0-10：mmaction2(SlowFast)-源码无死角解析（6）-模型构建总览
江南才尽江南山，年少无知年少狂！
08-12 2805
以下链接是个人关于mmaction2(SlowFast-动作识别) 所有见解，如有错误欢迎大家指出，我会第一时间纠正。有兴趣的朋友可以加微信：a944284742相互讨论技术。若是帮助到了你什么，一定要记得点赞！因为这是对我最大的鼓励。 动作识别0-00：mmaction2(SlowFast)-目录-史上最新无死角讲解 前言 通过前面的博客，我们已经知道训练模型的总体思路，以及数据集加载的整体过程，并且知道我们拿到的数据是什么。接下来我们就是要对模型进行分析。分析其是如何构建的，如何训练的，loss是如何计
slowfast 跑多卡的时候遇到问题
keyanxiaocaicai的专栏
11-24 336
问题描述： 在57 服务器上跑： python tools/run_net.py --cfg configs/Kinetics/X3D_XS.yaml NUM_GPUS 4 TRAIN.BATCH_SIZE 8 SOLVER.BASE_LR 0.0125 DATA.PATH_TO_DATA_DIR /mnt/data/geguojing/Heart_data/annotations20201112 目前这个问题没有解决，57 的服务器一直在被占用。调试不了 185 上的出现的错误是： ...
动作识别mmaction2使用方法和多视频帧图推理
最新发布
jin__9981的博客
01-29 1077
MMAction2 是一款基于 PyTorch 开发的行为识别开源工具包，是open-mmlab项目的一个子项目。
西工大-计算机学院-复试-面试题目
lixiaohulife的博客
03-18 5930
仅供参考，可以自行补充 计算机学院一系 1.操作系统的目录(分别学了哪些东西),什么是进程，进程调度是调度谁，进程的工作状态 （1）计算机操作系统的目录：全书共分12章 处理器管理，进程调度，存储管理，设备管理，文件管理 第一章为操作系统引论，介绍了OS的发展、传统OS和现代OS的特征及功能； 第二和第三章深入阐述了进程和线程管理、进程同步、处理机调度和死锁； 第四和第五章对连续和离散存储器管理方式及虚拟存储器进行了介绍； 第六章自下而上地对I/O系统的各个层次做了较为系统的阐述； 第七和第八章介绍了文件
机器学习面试系列四
Learn_ZhangK的博客
11-20 1738
61.说说梯度下降法   @LeftNotEasy，本题解析来源：http://www.cnblogs.com/LeftNotEasy/archive/2010/12/05/mathmatic_in_machine_learning_1_regression_and_gradient_descent.html   下面是一个典型的机器学习的过程，首先给出一个输入数据，我们的算法会通过一
动作识别0-01：mmaction2(SlowFast)-资源下载（前奏准备）
江南才尽江南山，年少无知年少狂！
07-23 4764
以下链接是个人关于mmaction2(SlowFast-动作识别) 所有见解，如有错误欢迎大家指出，我会第一时间纠正。有兴趣的朋友可以加微信：a944284742相互讨论技术。若是帮助到了你什么，一定要记得点赞！因为这是对我最大的鼓励。 动作识别0-00：mmaction2(SlowFast)-目录-史上最新无死角讲解 mmaction2 开源github：https://github.com/open-mmlab/mmaction2 论文地址 论文名称：SlowFast Networks for Vide
SlowFast 论文解读以及记录
qq_39598271的博客
05-11 667
引言 自然图片：I(x,y)I(x,y)I(x,y)，x与y同等对待 视频信号：I(x,y,t)I(x,y,t)I(x,y,t)，x，y与t不能同等对待 常规的三维时空卷积是不合理的 时间和空间分别训练的策略——研究进展缓慢 动作motion的语义划分策略——研究更新缓慢 某些motion要比先前的识别目标更加快（clap，wave，shake）——需要重视快动作 Fast Pathway——捕获动作的快速变化、高帧率、参数量小 Slow Pathway——捕获空间域和语义信息、低帧率、参数量大
mmaction2空间行为识别slowfast【demo_spatiotemporal_det】代码详解
重剑无锋博客
03-23 197
从中间帧开始，隔8帧（args.predict_stepsize）帧取一次，到帧数减去中间帧。我们获取到slowfast的配置参数。一共64帧隔2s取一帧，一共32帧。把刚才的帧作为中间帧进行检测。
mmaction2-时空动作检测slowfast笔记
qq_42987894的博客
08-20 2142
识别
FastFCN-master_FastFCN：_fast-rcnn_wet2cf_源码
10-04
python环境下的fcn模型，可供大家下载使用
Android源码：高仿360手机卫士2015-04
04-16
Android源码：高仿360手机卫士。深度访问手机资源权限。 更新于2015年4月，有需要的朋友可以参考下。 0分，方便没有分的朋友。
slowfast-network 何凯明源码
03-06
slow-fast network 何凯明最新论文源码，内含部分注释
java源码包---java 源码 大量 实例
04-18
Applet钢琴模拟程序java源码 2个目标文件，提供基本的音乐编辑功能。编辑音乐软件的朋友，这款实例会对你有所帮助。 Calendar万年历 1个目标文件 EJB 模拟银行ATM流程及操作源代码 6个目标文件，EJB来模拟银行ATM...
cocos2d-x+lua游戏demo源码
07-24
cocos2d-x+lua游戏demo源码 xcode工程
Grad-CAM可视化
bairw_Bella的博客
11-26 9266
GradCam可视化ResNet50范例 import argparse import cv2 import numpy as np import torch from torch.autograd import Function from torchvision import models class FeatureExtractor(): """ Class for extracting activations and registering gradients from tar
pytorch mmcv工程之卷积层定义
bairw_Bella的博客
03-17 4639
MMCV卷积层定义 Introduction MMCV is a foundational python library for computer vision research and supports many research projects as below: MMDetection: Detection toolbox and benchmark MMDetection3D: General 3D object detection toolbox and benchmark MMSegme
MMCV 加载模型配置文件
bairw_Bella的博客
04-12 2172
mmcv.Config 类 支持的配置文件类型包括：'.py', '.json', '.yaml', '.yml' 1，读取模型配置文件 from mmcv import Config cfg = Config.fromfile(args.config) 2, Config.fromfile() def fromfile(filename, use_predefined_variables=True, i
pytorch mmcv Resnet3D网络源码解读
bairw_Bella的博客
10-29 1779
Resnet3D网络 使用mmcv 构建残差块 1，模型初始化中构建残差块的基础模块，适用于浅层ResNet18,ResNet34 1，根据是否使用膨胀的3D卷积，确定卷积核的尺寸，padding和步长 if self.inflate: conv1_kernel_size = (3, 3, 3) conv1_padding = (1, dilation, dilation) conv2_kernel_size = (3,
slowfast源码复现
07-30
对于复现slowfast源码，可以按照以下步骤进行操作： 1. 数据准备：首先需要准备用于训练的数据集。可以使用已有的动作识别数据集，或者根据自己的需求创建一个数据集。数据集应包含视频文件和对应的标签。 2. 模型搭建：根据slowfast模型的论文或源码，搭建相应的网络结构。可以使用深度学习框架如PyTorch或TensorFlow来实现模型。 3. 训练模型：使用准备好的数据集对模型进行训练。可以使用GPU加速来提高训练速度和性能。训练过程中可以使用一些优化技巧，如学习率调整、数据增强等。 4. 导出模型：训练完成后，将模型导出为ONNX格式，以便后续的推理部署。可以使用深度学习框架提供的导出功能或者使用第三方库进行导出。 5. 推理部署：使用TensorRT等工具将ONNX模型进行优化和部署，以提高推理速度和效率。可以根据具体需求进行配置和调优。 6. 验证和测试：使用准备好的测试集对部署好的模型进行验证和测试，评估模型的性能和准确率。 需要注意的是，复现slowfast源码需要具备一定的深度学习和计算机视觉的知识，并且需要有相应的硬件和软件环境支持。同时，源码复现可能会遇到一些问题和挑战，需要进行适当的调试和优化。

“相关推荐”对你有帮助么？

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论 1
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值