PSMNet debug(三)

最新推荐文章于 2024-10-18 10:41:41 发布

_coconan_

最新推荐文章于 2024-10-18 10:41:41 发布

阅读量1k

点赞数

本文链接：https://blog.youkuaiyun.com/lzglzj20100700/article/details/84978815

版权

本文详细分析了PSMNet模型中stackhourglass模块的代码实现，包括convbn_3d函数、PSMNet类的初始化、以及hourglass模块的结构和前向传播过程。通过打印输出，展示了不同层的输出尺寸变化，揭示了网络中信息流程和特征处理的过程。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

stackhourglass.py

def convbn_3d(in_planes, out_planes, kernel_size, stride, pad):

return nn.Sequential(nn.Conv3d(in_planes, out_planes, kernel_size=kernel_size, padding=pad, stride=stride,bias=False),
nn.BatchNorm3d(out_planes))

class PSMNet(nn.Module):
def __init__(self, maxdisp):

self.classif1 = nn.Sequential(convbn_3d(32, 32, 3, 1, 1),
nn.ReLU(inplace=True),
nn.Conv3d(32, 1, kernel_size=3, padding=1, stride=1,bias=False))

self.classif2 = nn.Sequential(convbn_3d(32, 32, 3, 1, 1),
nn.ReLU(inplace=True),
nn.Conv3d(32, 1, kernel_size=3, padding=1, stride=1,bias=False))

self.classif3 = nn.Sequential(convbn_3d(32, 32, 3, 1, 1),
nn.ReLU(inplace=True),
nn.Conv3d(32, 1, kernel_size=3, padding=1, stride=1,bias=False))

print('stackhourglass classif1')
print(out1.size()[0])
print(out1.size()[1])
print(out1.size()[2])
print(out1.size()[3])
print(out1.size()[4])
cost1 = self.classif1(out1)
print('stackhourglass classif1 cost1')
print(cost1.size()[0])
print(cost1.size()[1]) #通道2为32，classif1 (nn.Conv3d(32, 1)处理后变为1。
print(cost1.size()[2])
print(cost1.size()[3])
print(cost1.size()[4])

打印结果：

stackhourglass classif1
1
32
48
96
312
stackhourglass classif1 cost1
1
1
48
96
312

cost2 = self.classif2(out2) + cost1
print('stackhourglass cost2')
print(cost2.size()[0])
print(c