FCOS

 

torch:

weight 128m 38.7

https://github.com/tianzhi0549/FCOS

 

pip install git+https://github.com/tianzhi0549/FCOS.git

精度有所提高，1070 速度170ms

只有部分代碼，沒有權重，權重參考上面的

https://github.com/feifeiwei/FCOS.pytorch

 

各種都有：也有訓練

https://github.com/Lausannen/NAS-FCOS

 

代碼解析：第一層就是7*7的卷積核

https://github.com/feifeiwei/FCOS.pytorch/blob/master/models/layers.py

 

70*100ms

 

# -*- coding: utf-8 -*-
import time

import torch
import torch.nn as nn
import torch.nn.functional as F


class Bottleneck(nn.Module):
    expansion = 4

    def __init__(self, in_planes, planes, stride=1):
        super(Bottleneck, self).__init__()
        self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=1, bias=False)
        self.bn1 = nn.BatchNorm2d(planes)
        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(planes)
        self.conv3 = nn.Conv2d(planes, self.expansion * planes, kernel_size=1, bias=False)
        self.bn3 = nn.BatchNorm2d(self.expansion * planes)

        self.downsample = nn.Sequential()
        if stride != 1 or in_planes != self.expansion * planes:
            self.downsample = nn.Sequential(
                nn.Conv2d(in_planes, self.expansion * planes, kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(self.expansion * planes)
            )

    def forward(self, x):
        out = F.relu(self.bn1(self.conv1(x)))
        out = F.relu(self.bn2(self.conv2(out)))
        out = self.bn3(self.conv3(out))
        out += self.downsample(x)
        out = F.relu(out)
        return out


class FPN(nn.Module):
    def __init__(self, block, num_blocks):
        super(FPN, self).__init__()
        self.in_planes = 64

        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False)
        self.bn1 = nn.BatchNorm2d(64)

        # Bottom-up layers
        self.layer1 = self._make_layer(block, 64, num_blocks[0], stride=1)
        self.layer2 = self._make_layer(block, 128, num_blocks[1], stride=2)
        self.layer3 = self._make_layer(block, 256, num_blocks[2], stride=2)
        self.layer4 = self._make_layer(block, 512, num_blocks[3], stride=2)

        self.conv6 = nn.Conv2d(2048, 256, kernel_size=3, stride=2, padding=1)
        self.conv7 = nn.Conv2d(256, 256, kernel_size=3, stride=2, padding=1)

        # Lateral layers
        self.latlayer1 = nn.Conv2d(2048, 256, kernel_size=1, stride=1, padding=0)
        self.latlayer2 = nn.Conv2d(1024, 256, kernel_size=1, stride=1, padding=0)
        self.latlayer3 = nn.Conv2d(512, 256, kernel_size=1, stride=1, padding=0)

        # smooth layers
        self.smooth1 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1)
        self.smooth2 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1)
        self.smooth3 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1)
        self.smooth4 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1)
        self.smooth5 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1)

    def _make_layer(self, block, planes, num_blocks, stride):
        strides = [stride] + [1] * (num_blocks - 1)
        layers = []
        for stride in strides:
            layers.append(block(self.in_planes, planes, stride))
            self.in_planes = planes * block.expansion
        return nn.Sequential(*layers)

    def _upsample_add(self, x, y):
        '''Upsample and add two feature maps.
        Args:
          x: (Variable) top feature map to be upsampled.
          y: (Variable) lateral feature map.
        Returns:
          (Variable) added feature map.
        Note in PyTorch, when input size is odd, the upsampled feature map
        with `F.upsample(..., scale_factor=2, mode='nearest')`
        maybe not equal to the lateral feature map size.
        e.g.
        original input size: [N,_,15,15] ->
        conv2d feature map size: [N,_,8,8] ->
        upsampled feature map size: [N,_,16,16]
        So we choose bilinear upsample which supports arbitrary output sizes.
        '''
        _, _, H, W = y.size()
        x = nn.functional.interpolate(x, scale_factor=2, mode='bilinear', align_corners=True)
        return x + y

    def forward(self, x):
        # Bottom-up
        c1 = F.relu(self.bn1(self.conv1(x)))  # /2
        c1 = F.max_pool2d(c1, kernel_size=3, stride=2, padding=1)  # /4
        c2 = self.layer1(c1)
        c3 = self.layer2(c2)  # / 8, 512
        c4 = self.layer3(c3)  # / 16, 1024
        c5 = self.layer4(c4)  # / 32, 2048

        p6 = self.conv6(c5)  # / 64, 256
        p7 = self.conv7(F.relu(p6))  # /128, 256
        # Top-down
        p5 = self.latlayer1(c5)
        p4 = self._upsample_add(p5, self.latlayer2(c4))
        p3 = self._upsample_add(p4, self.latlayer3(c3))
        # smooth
        p3 = self.smooth1(p3)  # /8
        p4 = self.smooth1(p4)  # /16
        p5 = self.smooth1(p5)  # /32
        p6 = self.smooth1(p6)  # /64
        p7 = self.smooth1(p7)  # /128
        return p3, p4, p5, p6, p7


def FPN50():
    return FPN(Bottleneck, [3, 4, 6, 3])


if __name__ == "__main__":
    x = torch.randn(1, 3, 640, 640).cuda()
    b = Bottleneck(3, 64, 2).cuda()
    n = FPN50().cuda()

    for i in range(10):
        start=time.time()
        y = b(x)
        yy = n(x)
        print(f'time:{time.time()-start} x: {x.shape}')
        # print(f'y: {y.shape}')
        # for i in yy:
        #     print(i.shape)
#    print(f'fpn: {[i.shape for i in yy]}')