Spaces:
Running
Running
File size: 3,633 Bytes
1c1a0c5 |
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 99 100 101 102 103 |
"""
UNet with Depthwise Separable Convolutions for efficient lane segmentation.
"""
import torch
import torch.nn as nn
import torch.nn.functional as F
class DepthwiseSeparableConv(nn.Module):
"""
Depthwise Separable Convolution Block.
Args:
in_channels (int): Number of input channels
out_channels (int): Number of output channels
"""
def __init__(self, in_channels: int, out_channels: int):
super().__init__()
self.depthwise = nn.Conv2d(in_channels, in_channels, kernel_size=3, padding=1, groups=in_channels)
self.pointwise = nn.Conv2d(in_channels, out_channels, kernel_size=1)
self.bn = nn.BatchNorm2d(out_channels)
self.relu = nn.ReLU(inplace=True)
def forward(self, x: torch.Tensor) -> torch.Tensor:
x = self.depthwise(x)
x = self.pointwise(x)
x = self.bn(x)
x = self.relu(x)
return x
class UNetDepthwise(nn.Module):
"""
UNet architecture using Depthwise Separable Convolutions.
Args:
in_channels (int): Number of input channels
out_channels (int): Number of output channels
"""
def __init__(self, in_channels: int = 3, out_channels: int = 1):
super().__init__()
self.enc1 = DepthwiseSeparableConv(in_channels, 64)
self.pool1 = nn.MaxPool2d(2)
self.enc2 = DepthwiseSeparableConv(64, 128)
self.pool2 = nn.MaxPool2d(2)
self.enc3 = DepthwiseSeparableConv(128, 256)
self.pool3 = nn.MaxPool2d(2)
self.enc4 = DepthwiseSeparableConv(256, 512)
self.pool4 = nn.MaxPool2d(2)
self.bottleneck = DepthwiseSeparableConv(512, 1024)
self.up4 = nn.ConvTranspose2d(1024, 512, kernel_size=2, stride=2)
self.dec4 = DepthwiseSeparableConv(1024, 512)
self.up3 = nn.ConvTranspose2d(512, 256, kernel_size=2, stride=2)
self.dec3 = DepthwiseSeparableConv(512, 256)
self.up2 = nn.ConvTranspose2d(256, 128, kernel_size=2, stride=2)
self.dec2 = DepthwiseSeparableConv(256, 128)
self.up1 = nn.ConvTranspose2d(128, 64, kernel_size=2, stride=2)
self.dec1 = DepthwiseSeparableConv(128, 64)
self.final_conv = nn.Conv2d(64, out_channels, kernel_size=1)
def forward(self, x: torch.Tensor) -> torch.Tensor:
# Encoder
e1 = self.enc1(x)
p1 = self.pool1(e1)
e2 = self.enc2(p1)
p2 = self.pool2(e2)
e3 = self.enc3(p2)
p3 = self.pool3(e3)
e4 = self.enc4(p3)
p4 = self.pool4(e4)
# Bottleneck
b = self.bottleneck(p4)
# Decoder
u4 = self.up4(b)
d4 = self.dec4(torch.cat([u4, e4], dim=1))
u3 = self.up3(d4)
d3 = self.dec3(torch.cat([u3, e3], dim=1))
u2 = self.up2(d3)
d2 = self.dec2(torch.cat([u2, e2], dim=1))
u1 = self.up1(d2)
d1 = self.dec1(torch.cat([u1, e1], dim=1))
out = self.final_conv(d1)
return torch.sigmoid(out)
# --- Model Summary and FLOPs (Optional) ---
if __name__ == "__main__":
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = UNetDepthwise(in_channels=3, out_channels=1).to(device)
dummy_input = torch.randn(1, 3, 256, 256).to(device)
output = model(dummy_input)
print(f"Output shape: {output.shape}")
# Model summary
from torchinfo import summary
print(summary(model, input_size=(1, 3, 256, 256), device=device))
# FLOPs and parameters
from thop import profile
flops, params = profile(model, inputs=(dummy_input,))
print(f"FLOPs: {flops:,}")
print(f"Parameters: {params:,}") |