backend/python-services/predictors/unet.py

"""
UNet model implementation.
Matches the architecture from deep-starry/starry/unet/ for loading .chkpt checkpoints.
"""

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


class DoubleConv(nn.Module):
	"""(convolution => [BN] => ReLU) * 2"""

	def __init__(self, in_channels, out_channels, mid_channels=None):
		super().__init__()
		if not mid_channels:
			mid_channels = out_channels
		self.double_conv = nn.Sequential(
			nn.Conv2d(in_channels, mid_channels, kernel_size=3, padding=1),
			nn.BatchNorm2d(mid_channels),
			nn.ReLU(inplace=True),
			nn.Conv2d(mid_channels, out_channels, kernel_size=3, padding=1),
			nn.BatchNorm2d(out_channels),
			nn.ReLU(inplace=True),
		)

	def forward(self, x):
		return self.double_conv(x)


class Down(nn.Module):
	"""Downscaling with maxpool then double conv"""

	def __init__(self, in_channels, out_channels):
		super().__init__()
		self.maxpool_conv = nn.Sequential(
			nn.MaxPool2d(2),
			DoubleConv(in_channels, out_channels)
		)

	def forward(self, x):
		return self.maxpool_conv(x)


class Up(nn.Module):
	"""Upscaling then double conv"""

	def __init__(self, in_channels, out_channels, bilinear=True):
		super().__init__()
		if bilinear:
			self.up = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True)
			self.conv = DoubleConv(in_channels, out_channels, in_channels // 2)
		else:
			self.up = nn.ConvTranspose2d(in_channels, in_channels // 2, kernel_size=2, stride=2)
			self.conv = DoubleConv(in_channels, out_channels)

	def forward(self, x1, x2):
		x1 = self.up(x1)
		diffY = x2.size()[2] - x1.size()[2]
		diffX = x2.size()[3] - x1.size()[3]
		x1 = F.pad(x1, [diffX // 2, diffX - diffX // 2,
						diffY // 2, diffY - diffY // 2])
		x = torch.cat([x2, x1], dim=1)
		return self.conv(x)


class OutConv(nn.Module):
	def __init__(self, in_channels, out_channels):
		super().__init__()
		self.conv = nn.Conv2d(in_channels, out_channels, kernel_size=1)

	def forward(self, x):
		return self.conv(x)


class UNet(nn.Module):
	def __init__(self, n_channels, n_classes, classify_out=True, bilinear=True, depth=4, init_width=64):
		super().__init__()
		self.n_channels = n_channels
		self.n_classes = n_classes
		self.classify_out = classify_out
		self.depth = depth
		factor = 2 if bilinear else 1

		self.inc = DoubleConv(n_channels, init_width)
		self.outc = OutConv(init_width, n_classes)

		downs = []
		ups = []

		for d in range(depth):
			ic = init_width * (2 ** d)
			oc = ic * 2
			if d == depth - 1:
				oc //= factor
			downs.append(Down(ic, oc))

		for d in range(depth):
			ic = init_width * (2 ** (depth - d))
			oc = ic // 2
			if d < depth - 1:
				oc //= factor
			ups.append(Up(ic, oc, bilinear))

		self.downs = nn.ModuleList(modules=downs)
		self.ups = nn.ModuleList(modules=ups)

	def forward(self, input):
		xs = []
		x = self.inc(input)

		for down in self.downs:
			xs.append(x)
			x = down(x)

		xs.reverse()

		for i, up in enumerate(self.ups):
			xi = xs[i]
			x = up(x, xi)

		if not self.classify_out:
			return x

		logits = self.outc(x)
		return logits