Spaces:
Sleeping
Sleeping
File size: 4,330 Bytes
201936b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 |
"""Code from : https://debuggercafe.com/implementing-resnet18-in-pytorch-from-scratch/"""
import torch.nn as nn
import torch
from torchvision.ops import RoIPool
from torch import Tensor
from typing import Type
class BasicBlock(nn.Module):
def __init__(self, in_channels: int,out_channels: int,stride: int = 1,expansion: int = 1,downsample: nn.Module = None) -> None:
super(BasicBlock, self).__init__()
# Multiplicative factor for the subsequent conv2d layer's output channels.
# It is 1 for ResNet18 and ResNet34.
self.expansion = expansion
self.downsample = downsample
self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=stride, padding=1,bias=False)
self.bn1 = nn.BatchNorm2d(out_channels)
self.relu = nn.ReLU(inplace=True)
self.conv2 = nn.Conv2d(out_channels, out_channels*self.expansion, kernel_size=3, padding=1,bias=False)
self.bn2 = nn.BatchNorm2d(out_channels*self.expansion)
def forward(self, x: Tensor) -> Tensor:
identity = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
if self.downsample is not None:
identity = self.downsample(x)
out += identity
out = self.relu(out)
return out
class ResNet18(nn.Module):
def __init__(self, img_channels: int,num_layers: int,block: Type[BasicBlock],num_classes: int = 1000) -> None:
super(ResNet18, self).__init__()
if num_layers == 18:
# The following `layers` list defines the number of `BasicBlock`
# to use to build the network and how many basic blocks to stack
# together.
layers = [2, 2, 2, 2]
self.expansion = 1
self.in_channels = 64
# All ResNets (18 to 152) contain a Conv2d => BN => ReLU for the first
# three layers. Here, kernel size is 7.
self.conv1 = nn.Conv2d(in_channels=img_channels,out_channels=self.in_channels,kernel_size=7, stride=2,padding=3,bias=False)
self.bn1 = nn.BatchNorm2d(self.in_channels)
self.relu = nn.ReLU(inplace=True)
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
self.layer1 = self._make_layer(block, 64, layers[0])
self.layer2 = self._make_layer(block, 128, layers[1], stride=2)
self.layer3 = self._make_layer(block, 256, layers[2], stride=2)
self.layer4 = self._make_layer(block, 512, layers[3], stride=2)
self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
self.fc = nn.Linear(512*self.expansion, num_classes)
def _make_layer(self, block: Type[BasicBlock],out_channels: int,blocks: int,stride: int = 1) -> nn.Sequential:
downsample = None
if stride != 1:
"""
This should pass from `layer2` to `layer4` or
when building ResNets50 and above. Section 3.3 of the paper
Deep Residual Learning for Image Recognition
(https://arxiv.org/pdf/1512.03385v1.pdf).
"""
downsample = nn.Sequential(
nn.Conv2d(self.in_channels, out_channels*self.expansion,kernel_size=1,stride=stride,bias=False),
nn.BatchNorm2d(out_channels * self.expansion),
)
layers = []
layers.append(
block(
self.in_channels, out_channels, stride, self.expansion, downsample
)
)
self.in_channels = out_channels * self.expansion
for i in range(1, blocks):
layers.append(block(
self.in_channels,
out_channels,
expansion=self.expansion
))
return nn.Sequential(*layers)
def forward(self, x: Tensor) -> Tensor:
x = self.conv1(x)
x = self.bn1(x)
x = self.relu(x)
x = self.maxpool(x)
x = self.layer1(x)
x = self.layer2(x)
x = self.layer3(x)
x = self.layer4(x)
# The spatial dimension of the final layer's feature
# map should be (7, 7) for all ResNets.
#print('Dimensions of the last convolutional feature map: ', x.shape)
x = self.avgpool(x)
x = torch.flatten(x, 1)
#x = self.fc(x)
return x |