Merge remote-tracking branch 'origin/gan_lab' into gan_lab

codes/data/image_corruptor.py

@@ -71,8 +71,15 @@ class ImageCorruptor:
             # Large distortion blocks in part of an img, such as is used to mask out a face.
             pass
         elif 'lq_resampling' in aug:
-            # Bicubic LR->HR
+            # Random mode interpolation HR->LR->HR
-            pass
+            scale = 2
+            if 'lq_resampling4x' == aug:
+                scale = 4
+            interpolation_modes = [cv2.INTER_AREA, cv2.INTER_NEAREST, cv2.INTER_CUBIC, cv2.INTER_LINEAR, cv2.INTER_LANCZOS4]
+            mode = rand_int % len(interpolation_modes)
+            # Downsample first, then upsample using the random mode.
+            img = cv2.resize(img, dsize=(img.shape[1]//scale, img.shape[0]//scale), interpolation=cv2.INTER_NEAREST)
+            img = cv2.resize(img, dsize=(img.shape[1]*scale, img.shape[0]*scale), interpolation=mode)
         elif 'color_shift' in aug:
             # Color shift
             pass

codes/data/multiscale_dataset.py

@@ -13,6 +13,24 @@ import torchvision.transforms.functional as F
 from data.image_corruptor import ImageCorruptor
 
 
+# Selects the smallest dimension from the image and crops it randomly so the other dimension matches. The cropping
+# offset from center is chosen on a normal probability curve.
+def get_square_image(image):
+    h, w, _ = image.shape
+    if h == w:
+        return image
+    offset = max(min(np.random.normal(scale=.3), 1.0), -1.0)
+    if h > w:
+        diff = h - w
+        center = diff // 2
+        top = max(int(center + offset * (center - 2)), 0)
+        return image[top:top + w, :, :]
+    else:
+        diff = w - h
+        center = diff // 2
+        left = max(int(center + offset * (center - 2)), 0)
+        return image[:, left:left + h, :]
+
 class MultiScaleDataset(data.Dataset):
     def __init__(self, opt):
         super(MultiScaleDataset, self).__init__()
@@ -25,23 +43,6 @@ class MultiScaleDataset(data.Dataset):
         self.paths_hq, self.sizes_hq = util.get_image_paths(self.data_type, opt['paths'], [1 for _ in opt['paths']])
         self.corruptor = ImageCorruptor(opt)
 
-    # Selects the smallest dimension from the image and crops it randomly so the other dimension matches. The cropping
-    # offset from center is chosen on a normal probability curve.
-    def get_square_image(self, image):
-        h, w, _ = image.shape
-        if h == w:
-            return image
-        offset = max(min(np.random.normal(scale=.3), 1.0), -1.0)
-        if h > w:
-            diff = h - w
-            center = diff // 2
-            top = int(center + offset * (center - 2))
-            return image[top:top+w, :, :]
-        else:
-            diff = w - h
-            center = diff // 2
-            left = int(center + offset * (center - 2))
-            return image[:, left:left+h, :]
 
     def recursively_extract_patches(self, input_img, result_list, depth):
         if depth >= self.num_scales:
@@ -62,7 +63,7 @@ class MultiScaleDataset(data.Dataset):
         loaded_img = util.read_img(None, full_path, None)
         img_full1 = util.channel_convert(loaded_img.shape[2], 'RGB', [loaded_img])[0]
         img_full2 = util.augment([img_full1], True, True)[0]
-        img_full3 = self.get_square_image(img_full2)
+        img_full3 = get_square_image(img_full2)
         # This error crops up from time to time. I suspect an issue with util.read_img.
         if img_full3.shape[0] == 0 or img_full3.shape[1] == 0:
             print("Error with image: %s. Loaded image shape: %s" % (full_path,str(loaded_img.shape)), str(img_full1.shape), str(img_full2.shape), str(img_full3.shape))

codes/models/archs/discriminator_vgg_arch.py

@@ -513,3 +513,93 @@ class RefDiscriminatorVgg128(nn.Module):
 
         out = self.output_linears(torch.cat([fea, ref_vector], dim=1))
         return out
+
+
+class PsnrApproximator(nn.Module):
+    # input_img_factor = multiplier to support images over 128x128. Only certain factors are supported.
+    def __init__(self, nf, input_img_factor=1):
+        super(PsnrApproximator, self).__init__()
+
+        # [64, 128, 128]
+        self.fake_conv0_0 = nn.Conv2d(3, nf, 3, 1, 1, bias=True)
+        self.fake_conv0_1 = nn.Conv2d(nf, nf, 4, 2, 1, bias=False)
+        self.fake_bn0_1 = nn.BatchNorm2d(nf, affine=True)
+        # [64, 64, 64]
+        self.fake_conv1_0 = nn.Conv2d(nf, nf * 2, 3, 1, 1, bias=False)
+        self.fake_bn1_0 = nn.BatchNorm2d(nf * 2, affine=True)
+        self.fake_conv1_1 = nn.Conv2d(nf * 2, nf * 2, 4, 2, 1, bias=False)
+        self.fake_bn1_1 = nn.BatchNorm2d(nf * 2, affine=True)
+        # [128, 32, 32]
+        self.fake_conv2_0 = nn.Conv2d(nf * 2, nf * 4, 3, 1, 1, bias=False)
+        self.fake_bn2_0 = nn.BatchNorm2d(nf * 4, affine=True)
+        self.fake_conv2_1 = nn.Conv2d(nf * 4, nf * 4, 4, 2, 1, bias=False)
+        self.fake_bn2_1 = nn.BatchNorm2d(nf * 4, affine=True)
+
+        # [64, 128, 128]
+        self.real_conv0_0 = nn.Conv2d(3, nf, 3, 1, 1, bias=True)
+        self.real_conv0_1 = nn.Conv2d(nf, nf, 4, 2, 1, bias=False)
+        self.real_bn0_1 = nn.BatchNorm2d(nf, affine=True)
+        # [64, 64, 64]
+        self.real_conv1_0 = nn.Conv2d(nf, nf * 2, 3, 1, 1, bias=False)
+        self.real_bn1_0 = nn.BatchNorm2d(nf * 2, affine=True)
+        self.real_conv1_1 = nn.Conv2d(nf * 2, nf * 2, 4, 2, 1, bias=False)
+        self.real_bn1_1 = nn.BatchNorm2d(nf * 2, affine=True)
+        # [128, 32, 32]
+        self.real_conv2_0 = nn.Conv2d(nf * 2, nf * 4, 3, 1, 1, bias=False)
+        self.real_bn2_0 = nn.BatchNorm2d(nf * 4, affine=True)
+        self.real_conv2_1 = nn.Conv2d(nf * 4, nf * 4, 4, 2, 1, bias=False)
+        self.real_bn2_1 = nn.BatchNorm2d(nf * 4, affine=True)
+
+        # [512, 16, 16]
+        self.conv3_0 = nn.Conv2d(nf * 8, nf * 8, 3, 1, 1, bias=False)
+        self.bn3_0 = nn.BatchNorm2d(nf * 8, affine=True)
+        self.conv3_1 = nn.Conv2d(nf * 8, nf * 8, 4, 2, 1, bias=False)
+        self.bn3_1 = nn.BatchNorm2d(nf * 8, affine=True)
+        # [512, 8, 8]
+        self.conv4_0 = nn.Conv2d(nf * 8, nf * 8, 3, 1, 1, bias=False)
+        self.bn4_0 = nn.BatchNorm2d(nf * 8, affine=True)
+        self.conv4_1 = nn.Conv2d(nf * 8, nf * 8, 4, 2, 1, bias=False)
+        self.bn4_1 = nn.BatchNorm2d(nf * 8, affine=True)
+        final_nf = nf * 8
+
+        # activation function
+        self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True)
+        self.linear1 = nn.Linear(int(final_nf * 4 * input_img_factor * 4 * input_img_factor), 1024)
+        self.linear2 = nn.Linear(1024, 512)
+        self.linear3 = nn.Linear(512, 128)
+        self.linear4 = nn.Linear(128, 1)
+
+    def compute_body1(self, real):
+        fea = self.lrelu(self.real_conv0_0(real))
+        fea = self.lrelu(self.real_bn0_1(self.real_conv0_1(fea)))
+        fea = self.lrelu(self.real_bn1_0(self.real_conv1_0(fea)))
+        fea = self.lrelu(self.real_bn1_1(self.real_conv1_1(fea)))
+        fea = self.lrelu(self.real_bn2_0(self.real_conv2_0(fea)))
+        fea = self.lrelu(self.real_bn2_1(self.real_conv2_1(fea)))
+        return fea
+
+    def compute_body2(self, fake):
+        fea = self.lrelu(self.fake_conv0_0(fake))
+        fea = self.lrelu(self.fake_bn0_1(self.fake_conv0_1(fea)))
+        fea = self.lrelu(self.fake_bn1_0(self.fake_conv1_0(fea)))
+        fea = self.lrelu(self.fake_bn1_1(self.fake_conv1_1(fea)))
+        fea = self.lrelu(self.fake_bn2_0(self.fake_conv2_0(fea)))
+        fea = self.lrelu(self.fake_bn2_1(self.fake_conv2_1(fea)))
+        return fea
+
+    def forward(self, real, fake):
+        real_fea = checkpoint(self.compute_body1, real)
+        fake_fea = checkpoint(self.compute_body2, fake)
+        fea = torch.cat([real_fea, fake_fea], dim=1)
+
+        fea = self.lrelu(self.bn3_0(self.conv3_0(fea)))
+        fea = self.lrelu(self.bn3_1(self.conv3_1(fea)))
+        fea = self.lrelu(self.bn4_0(self.conv4_0(fea)))
+        fea = self.lrelu(self.bn4_1(self.conv4_1(fea)))
+
+        fea = fea.contiguous().view(fea.size(0), -1)
+        fea = self.lrelu(self.linear1(fea))
+        fea = self.lrelu(self.linear2(fea))
+        fea = self.lrelu(self.linear3(fea))
+        out = self.linear4(fea)
+        return out.squeeze()

codes/models/archs/srg2_classic.py

@@ -0,0 +1,421 @@
+import os
+
+import torch
+import torchvision
+from torch import nn
+import torch.nn.functional as F
+import functools
+from collections import OrderedDict
+
+from torch.nn import init
+
+from models.archs.arch_util import ConvBnLelu, ConvGnSilu
+from utils.util import checkpoint
+
+
+def initialize_weights(net_l, scale=1):
+    if not isinstance(net_l, list):
+        net_l = [net_l]
+    for net in net_l:
+        for m in net.modules():
+            if isinstance(m, nn.Conv2d) or isinstance(m, nn.Conv3d):
+                init.kaiming_normal_(m.weight, a=0, mode='fan_in')
+                m.weight.data *= scale  # for residual block
+                if m.bias is not None:
+                    m.bias.data.zero_()
+            elif isinstance(m, nn.Linear):
+                init.kaiming_normal_(m.weight, a=0, mode='fan_in')
+                m.weight.data *= scale
+                if m.bias is not None:
+                    m.bias.data.zero_()
+            elif isinstance(m, nn.BatchNorm2d):
+                init.constant_(m.weight, 1)
+                init.constant_(m.bias.data, 0.0)
+
+
+class AttentionNorm(nn.Module):
+    def __init__(self, group_size, accumulator_size=128):
+        super(AttentionNorm, self).__init__()
+        self.accumulator_desired_size = accumulator_size
+        self.group_size = group_size
+        # These are all tensors so that they get saved with the graph.
+        self.accumulator = nn.Parameter(torch.zeros(accumulator_size, group_size), requires_grad=False)
+        self.accumulator_index = nn.Parameter(torch.zeros(1, dtype=torch.long, device='cpu'), requires_grad=False)
+        self.accumulator_filled = nn.Parameter(torch.zeros(1, dtype=torch.bool, device='cpu'), requires_grad=False)
+
+    # Returns tensor of shape (group,) with a normalized mean across the accumulator in the range [0,1]. The intent
+    # is to divide your inputs by this value.
+    def compute_buffer_norm(self):
+        if self.accumulator_filled:
+            return torch.mean(self.accumulator, dim=0)
+        else:
+            return torch.ones(self.group_size, device=self.accumulator.device)
+
+    def add_norm_to_buffer(self, x):
+        flat = x.sum(dim=[0, 1, 2], keepdim=True)
+        norm = flat / torch.mean(flat)
+
+        # This often gets reset in GAN mode. We *never* want gradient accumulation in this parameter.
+        self.accumulator.requires_grad = False
+        self.accumulator[self.accumulator_index] = norm.detach()
+        self.accumulator_index += 1
+        if self.accumulator_index >= self.accumulator_desired_size:
+            self.accumulator_index *= 0
+            self.accumulator_filled |= True
+
+    # Input into forward is an attention tensor of shape (batch,width,height,groups)
+    def forward(self, x: torch.Tensor):
+        assert len(x.shape) == 4
+        # Push the accumulator to the right device on the first iteration.
+        if self.accumulator.device != x.device:
+            self.accumulator = self.accumulator.to(x.device)
+
+        self.add_norm_to_buffer(x)
+        norm = self.compute_buffer_norm()
+        x = x / norm
+
+        # Need to re-normalize x so that the groups dimension sum to 1, just like when it was fed in.
+        groups_sum = x.sum(dim=3, keepdim=True)
+        return x / groups_sum
+
+
+class BareConvSwitch(nn.Module):
+    """
+    Initializes the ConvSwitch.
+      initial_temperature: The initial softmax temperature of the attention mechanism. For training from scratch, this
+                           should be set to a high number, for example 30.
+      attention_norm:      If specified, the AttentionNorm layer applied immediately after Softmax.
+    """
+
+    def __init__(
+        self,
+        initial_temperature=1,
+        attention_norm=None
+    ):
+        super(BareConvSwitch, self).__init__()
+
+        self.softmax = nn.Softmax(dim=-1)
+        self.temperature = initial_temperature
+        self.attention_norm = attention_norm
+
+        initialize_weights(self)
+
+    def set_attention_temperature(self, temp):
+        self.temperature = temp
+
+    # SwitchedConv.forward takes these arguments;
+    # conv_group:      List of inputs (len=n) to the switch, each with shape (b,f,w,h)
+    # conv_attention:  Attention computation as an output from a conv layer, of shape (b,n,w,h). Before softmax
+    # output_attention_weights: If True, post-softmax attention weights are returned.
+    def forward(self, conv_group, conv_attention, output_attention_weights=False):
+        # Stack up the conv_group input first and permute it to (batch, width, height, filter, groups)
+        conv_outputs = torch.stack(conv_group, dim=0).permute(1, 3, 4, 2, 0)
+
+        conv_attention = conv_attention.permute(0, 2, 3, 1)
+        conv_attention = self.softmax(conv_attention / self.temperature)
+        if self.attention_norm:
+            conv_attention = self.attention_norm(conv_attention)
+
+        # conv_outputs shape:   (batch, width, height, filters, groups)
+        # conv_attention shape: (batch, width, height, groups)
+        # We want to format them so that we can matmul them together to produce:
+        # desired shape:        (batch, width, height, filters)
+        # Note: conv_attention will generally be cast to float32 regardless of the input type, so cast conv_outputs to
+        #       float32 as well to match it.
+        if self.training:
+            # Doing it all in one op is substantially faster - better for training.
+            attention_result = torch.einsum(
+                "...ij,...j->...i", [conv_outputs.float(), conv_attention]
+            )
+        else:
+            # eval_mode substantially reduces the GPU memory required to compute the attention result by performing the
+            # attention multiplications one at a time. This is probably necessary for large images and attention breadths.
+            attention_result = conv_outputs[:, :, :, :, 0] * conv_attention[:, :, :, 0].unsqueeze(dim=-1)
+            for i in range(1, conv_attention.shape[-1]):
+                attention_result += conv_outputs[:, :, :, :, i] * conv_attention[:, :, :, i].unsqueeze(dim=-1)
+
+        # Remember to shift the filters back into the expected slot.
+        if output_attention_weights:
+            return attention_result.permute(0, 3, 1, 2), conv_attention
+        else:
+            return attention_result.permute(0, 3, 1, 2)
+
+
+class MultiConvBlock(nn.Module):
+    def __init__(self, filters_in, filters_mid, filters_out, kernel_size, depth, scale_init=1.0, norm=False, weight_init_factor=1):
+        assert depth >= 2
+        super(MultiConvBlock, self).__init__()
+        self.noise_scale = nn.Parameter(torch.full((1,), fill_value=.01))
+        self.bnconvs = nn.ModuleList([ConvBnLelu(filters_in, filters_mid, kernel_size, norm=norm, bias=False, weight_init_factor=weight_init_factor)] +
+                                     [ConvBnLelu(filters_mid, filters_mid, kernel_size, norm=norm, bias=False, weight_init_factor=weight_init_factor) for i in range(depth - 2)] +
+                                     [ConvBnLelu(filters_mid, filters_out, kernel_size, activation=False, norm=False, bias=False, weight_init_factor=weight_init_factor)])
+        self.scale = nn.Parameter(torch.full((1,), fill_value=scale_init))
+        self.bias = nn.Parameter(torch.zeros(1), requires_grad=False)
+
+    def forward(self, x, noise=None):
+        if noise is not None:
+            noise = noise * self.noise_scale
+            x = x + noise
+        for m in self.bnconvs:
+            x = m.forward(x)
+        return x * self.scale + self.bias
+
+
+# VGG-style layer with Conv(stride2)->BN->Activation->Conv->BN->Activation
+# Doubles the input filter count.
+class HalvingProcessingBlock(nn.Module):
+    def __init__(self, filters):
+        super(HalvingProcessingBlock, self).__init__()
+        self.bnconv1 = ConvGnSilu(filters, filters * 2, stride=2, norm=False, bias=False)
+        self.bnconv2 = ConvGnSilu(filters * 2, filters * 2, norm=True, bias=False)
+
+    def forward(self, x):
+        x = self.bnconv1(x)
+        return self.bnconv2(x)
+
+
+# Block that upsamples 2x and reduces incoming filters by 2x. It preserves structure by taking a passthrough feed
+# along with the feature representation.
+class ExpansionBlock(nn.Module):
+    def __init__(self, filters_in, filters_out=None, block=ConvGnSilu):
+        super(ExpansionBlock, self).__init__()
+        if filters_out is None:
+            filters_out = filters_in // 2
+        self.decimate = block(filters_in, filters_out, kernel_size=1, bias=False, activation=False, norm=True)
+        self.process_passthrough = block(filters_out, filters_out, kernel_size=3, bias=True, activation=False, norm=True)
+        self.conjoin = block(filters_out*2, filters_out, kernel_size=3, bias=False, activation=True, norm=False)
+        self.process = block(filters_out, filters_out, kernel_size=3, bias=False, activation=True, norm=True)
+
+    # input is the feature signal with shape  (b, f, w, h)
+    # passthrough is the structure signal with shape (b, f/2, w*2, h*2)
+    # output is conjoined upsample with shape (b, f/2, w*2, h*2)
+    def forward(self, input, passthrough):
+        x = F.interpolate(input, scale_factor=2, mode="nearest")
+        x = self.decimate(x)
+        p = self.process_passthrough(passthrough)
+        x = self.conjoin(torch.cat([x, p], dim=1))
+        return self.process(x)
+
+
+# This is a classic u-net architecture with the goal of assigning each individual pixel an individual transform
+# switching set.
+class ConvBasisMultiplexer(nn.Module):
+    def __init__(self, input_channels, base_filters, reductions, processing_depth, multiplexer_channels, use_gn=True):
+        super(ConvBasisMultiplexer, self).__init__()
+        self.filter_conv = ConvGnSilu(input_channels, base_filters, bias=True)
+        self.reduction_blocks = nn.ModuleList([HalvingProcessingBlock(base_filters * 2 ** i) for i in range(reductions)])
+        reduction_filters = base_filters * 2 ** reductions
+        self.processing_blocks = nn.Sequential(OrderedDict([('block%i' % (i,), ConvGnSilu(reduction_filters, reduction_filters, bias=False)) for i in range(processing_depth)]))
+        self.expansion_blocks = nn.ModuleList([ExpansionBlock(reduction_filters // (2 ** i)) for i in range(reductions)])
+
+        gap = base_filters - multiplexer_channels
+        cbl1_out = ((base_filters - (gap // 2)) // 4) * 4   # Must be multiples of 4 to use with group norm.
+        self.cbl1 = ConvGnSilu(base_filters, cbl1_out, norm=use_gn, bias=False, num_groups=4)
+        cbl2_out = ((base_filters - (3 * gap // 4)) // 4) * 4
+        self.cbl2 = ConvGnSilu(cbl1_out, cbl2_out, norm=use_gn, bias=False, num_groups=4)
+        self.cbl3 = ConvGnSilu(cbl2_out, multiplexer_channels, bias=True, norm=False)
+
+    def forward(self, x):
+        x = self.filter_conv(x)
+        reduction_identities = []
+        for b in self.reduction_blocks:
+            reduction_identities.append(x)
+            x = b(x)
+        x = self.processing_blocks(x)
+        for i, b in enumerate(self.expansion_blocks):
+            x = b(x, reduction_identities[-i - 1])
+
+        x = self.cbl1(x)
+        x = self.cbl2(x)
+        x = self.cbl3(x)
+        return x
+
+
+class ConfigurableSwitchComputer(nn.Module):
+    def __init__(self, base_filters, multiplexer_net, pre_transform_block, transform_block, transform_count, init_temp=20,
+                 add_scalable_noise_to_transforms=False):
+        super(ConfigurableSwitchComputer, self).__init__()
+
+        tc = transform_count
+        self.multiplexer = multiplexer_net(tc)
+
+        self.pre_transform = pre_transform_block()
+        self.transforms = nn.ModuleList([transform_block() for _ in range(transform_count)])
+        self.add_noise = add_scalable_noise_to_transforms
+        self.noise_scale = nn.Parameter(torch.full((1,), float(1e-3)))
+
+        # And the switch itself, including learned scalars
+        self.switch = BareConvSwitch(initial_temperature=init_temp, attention_norm=AttentionNorm(transform_count, accumulator_size=16 * transform_count))
+        self.switch_scale = nn.Parameter(torch.full((1,), float(1)))
+        self.post_switch_conv = ConvBnLelu(base_filters, base_filters, norm=False, bias=True)
+        # The post_switch_conv gets a low scale initially. The network can decide to magnify it (or not)
+        # depending on its needs.
+        self.psc_scale = nn.Parameter(torch.full((1,), float(.1)))
+
+    def forward(self, x, output_attention_weights=True):
+        identity = x
+        if self.add_noise:
+            rand_feature = torch.randn_like(x) * self.noise_scale
+            x = x + rand_feature
+
+        x = self.pre_transform(x)
+        xformed = [t.forward(x) for t in self.transforms]
+
+        m = self.multiplexer(identity)
+
+        outputs, attention = self.switch(xformed, m, True)
+        outputs = identity + outputs * self.switch_scale
+        outputs = outputs + self.post_switch_conv(outputs) * self.psc_scale
+        if output_attention_weights:
+            return outputs, attention
+        else:
+            return outputs
+
+    def set_temperature(self, temp):
+        self.switch.set_attention_temperature(temp)
+
+
+def compute_attention_specificity(att_weights, topk=3):
+    att = att_weights.detach()
+    vals, indices = torch.topk(att, topk, dim=-1)
+    avg = torch.sum(vals, dim=-1)
+    avg = avg.flatten().mean()
+    return avg.item(), indices.flatten().detach()
+
+
+# Copied from torchvision.utils.save_image. Allows specifying pixel format.
+def save_image(tensor, fp, nrow=8, padding=2,
+               normalize=False, range=None, scale_each=False, pad_value=0, format=None, pix_format=None):
+    from PIL import Image
+    grid = torchvision.utils.make_grid(tensor, nrow=nrow, padding=padding, pad_value=pad_value,
+                                       normalize=normalize, range=range, scale_each=scale_each)
+    # Add 0.5 after unnormalizing to [0, 255] to round to nearest integer
+    ndarr = grid.mul(255).add_(0.5).clamp_(0, 255).permute(1, 2, 0).to('cpu', torch.uint8).numpy()
+    im = Image.fromarray(ndarr, mode=pix_format).convert('RGB')
+    im.save(fp, format=format)
+
+
+def save_attention_to_image(folder, attention_out, attention_size, step, fname_part="map", l_mult=1.0):
+    magnitude, indices = torch.topk(attention_out, 1, dim=-1)
+    magnitude = magnitude.squeeze(3)
+    indices = indices.squeeze(3)
+    # indices is an integer tensor (b,w,h) where values are on the range [0,attention_size]
+    # magnitude is a float tensor (b,w,h) [0,1] representing the magnitude of that attention.
+    # Use HSV colorspace to show this. Hue is mapped to the indices, Lightness is mapped to intensity,
+    # Saturation is left fixed.
+    hue = indices.float() / attention_size
+    saturation = torch.full_like(hue, .8)
+    value = magnitude * l_mult
+    hsv_img = torch.stack([hue, saturation, value], dim=1)
+
+    output_path=os.path.join(folder, "attention_maps", fname_part)
+    os.makedirs(output_path, exist_ok=True)
+    save_image(hsv_img, os.path.join(output_path, "attention_map_%i.png" % (step,)), pix_format="HSV")
+
+
+def save_attention_to_image_rgb(output_folder, attention_out, attention_size, file_prefix, step, cmap_discrete_name='viridis'):
+    magnitude, indices = torch.topk(attention_out, 3, dim=-1)
+    magnitude = magnitude.cpu()
+    indices = indices.cpu()
+    magnitude /= torch.max(torch.abs(torch.min(magnitude)), torch.abs(torch.max(magnitude)))
+    colormap = cm.get_cmap(cmap_discrete_name, attention_size)
+    colormap_mag = cm.get_cmap(cmap_discrete_name)
+    os.makedirs(os.path.join(output_folder), exist_ok=True)
+    for i in range(3):
+        img = torch.tensor(colormap(indices[:,:,:,i].detach().numpy()))
+        img = img.permute((0, 3, 1, 2))
+        save_image(img, os.path.join(output_folder, file_prefix + "_%i_%s.png" % (step, "rgb_%i" % (i,))), pix_format="RGBA")
+
+        mag_image = torch.tensor(colormap_mag(magnitude[:,:,:,i].detach().numpy()))
+        mag_image = mag_image.permute((0, 3, 1, 2))
+        save_image(mag_image, os.path.join(output_folder, file_prefix + "_%i_%s.png" % (step, "mag_%i" % (i,))), pix_format="RGBA")
+
+
+class ConfigurableSwitchedResidualGenerator2(nn.Module):
+    def __init__(self, switch_depth, switch_filters, switch_reductions, switch_processing_layers, trans_counts, trans_kernel_sizes,
+                 trans_layers, transformation_filters, initial_temp=20, final_temperature_step=50000, heightened_temp_min=1,
+                 heightened_final_step=50000, upsample_factor=1,
+                 add_scalable_noise_to_transforms=False):
+        super(ConfigurableSwitchedResidualGenerator2, self).__init__()
+        switches = []
+        self.initial_conv = ConvBnLelu(3, transformation_filters, norm=False, activation=False, bias=True)
+        self.upconv1 = ConvBnLelu(transformation_filters, transformation_filters, norm=False, bias=True)
+        self.upconv2 = ConvBnLelu(transformation_filters, transformation_filters, norm=False, bias=True)
+        self.hr_conv = ConvBnLelu(transformation_filters, transformation_filters, norm=False, bias=True)
+        self.final_conv = ConvBnLelu(transformation_filters, 3, norm=False, activation=False, bias=True)
+        for _ in range(switch_depth):
+            multiplx_fn = functools.partial(ConvBasisMultiplexer, transformation_filters, switch_filters, switch_reductions, switch_processing_layers, trans_counts)
+            pretransform_fn = functools.partial(ConvBnLelu, transformation_filters, transformation_filters, norm=False, bias=False, weight_init_factor=.1)
+            transform_fn = functools.partial(MultiConvBlock, transformation_filters, int(transformation_filters * 1.5), transformation_filters, kernel_size=trans_kernel_sizes, depth=trans_layers, weight_init_factor=.1)
+            switches.append(ConfigurableSwitchComputer(transformation_filters, multiplx_fn,
+                                                       pre_transform_block=pretransform_fn, transform_block=transform_fn,
+                                                       transform_count=trans_counts, init_temp=initial_temp,
+                                                       add_scalable_noise_to_transforms=add_scalable_noise_to_transforms))
+
+        self.switches = nn.ModuleList(switches)
+        self.transformation_counts = trans_counts
+        self.init_temperature = initial_temp
+        self.final_temperature_step = final_temperature_step
+        self.heightened_temp_min = heightened_temp_min
+        self.heightened_final_step = heightened_final_step
+        self.attentions = None
+        self.upsample_factor = upsample_factor
+        assert self.upsample_factor == 2 or self.upsample_factor == 4
+
+    def forward(self, x):
+        x = self.initial_conv(x)
+
+        self.attentions = []
+        for i, sw in enumerate(self.switches):
+            x, att = checkpoint(sw, x)
+            self.attentions.append(att)
+
+        x = self.upconv1(F.interpolate(x, scale_factor=2, mode="nearest"))
+        if self.upsample_factor > 2:
+            x = F.interpolate(x, scale_factor=2, mode="nearest")
+        x = self.upconv2(x)
+        x = self.final_conv(self.hr_conv(x))
+        return x
+
+    def set_temperature(self, temp):
+        [sw.set_temperature(temp) for sw in self.switches]
+
+    def update_for_step(self, step, experiments_path='.'):
+        if self.attentions:
+            temp = max(1,
+                1 + self.init_temperature * (self.final_temperature_step - step) / self.final_temperature_step)
+            if temp == 1 and self.heightened_final_step and step > self.final_temperature_step and \
+                    self.heightened_final_step != 1:
+                # Once the temperature passes (1) it enters an inverted curve to match the linear curve from above.
+                # without this, the attention specificity "spikes" incredibly fast in the last few iterations.
+                h_steps_total = self.heightened_final_step - self.final_temperature_step
+                h_steps_current = min(step - self.final_temperature_step, h_steps_total)
+                # The "gap" will represent the steps that need to be traveled as a linear function.
+                h_gap = 1 / self.heightened_temp_min
+                temp = h_gap * h_steps_current / h_steps_total
+                # Invert temperature to represent reality on this side of the curve
+                temp = 1 / temp
+            self.set_temperature(temp)
+            if step % 50 == 0:
+                [save_attention_to_image(experiments_path, self.attentions[i], self.transformation_counts, step, "a%i" % (i+1,), l_mult=10) for i in  range(len(self.attentions))]
+
+    def get_debug_values(self, step, net_name):
+        temp = self.switches[0].switch.temperature
+        mean_hists = [compute_attention_specificity(att, 2) for att in self.attentions]
+        means = [i[0] for i in mean_hists]
+        hists = [i[1].clone().detach().cpu().flatten() for i in mean_hists]
+        val = {"switch_temperature": temp}
+        for i in range(len(means)):
+            val["switch_%i_specificity" % (i,)] = means[i]
+            val["switch_%i_histogram" % (i,)] = hists[i]
+        return val
+
+
+class Interpolate(nn.Module):
+    def __init__(self, factor):
+        super(Interpolate, self).__init__()
+        self.factor = factor
+
+    def forward(self, x):
+        return F.interpolate(x, scale_factor=self.factor)
+

codes/models/networks.py

@@ -18,6 +18,7 @@ import models.archs.feature_arch as feature_arch
 import models.archs.panet.panet as panet
 import models.archs.rcan as rcan
 import models.archs.ChainedEmbeddingGen as chained
+from models.archs import srg2_classic
 from models.archs.teco_resgen import TecoGen
 
 logger = logging.getLogger('base')
@@ -64,6 +65,15 @@ def define_G(opt, net_key='network_G', scale=None):
                                                                       initial_temp=opt_net['temperature'], final_temperature_step=opt_net['temperature_final_step'],
                                                                       heightened_temp_min=opt_net['heightened_temp_min'], heightened_final_step=opt_net['heightened_final_step'],
                                                                       upsample_factor=scale, add_scalable_noise_to_transforms=opt_net['add_noise'])
+    elif which_model == "srg2classic":
+        netG = srg2_classic.ConfigurableSwitchedResidualGenerator2(switch_depth=opt_net['switch_depth'], switch_filters=opt_net['switch_filters'],
+                                                                      switch_reductions=opt_net['switch_reductions'],
+                                                                      switch_processing_layers=opt_net['switch_processing_layers'], trans_counts=opt_net['trans_counts'],
+                                                                      trans_kernel_sizes=opt_net['trans_kernel_sizes'], trans_layers=opt_net['trans_layers'],
+                                                                      transformation_filters=opt_net['transformation_filters'],
+                                                                      initial_temp=opt_net['temperature'], final_temperature_step=opt_net['temperature_final_step'],
+                                                                      heightened_temp_min=opt_net['heightened_temp_min'], heightened_final_step=opt_net['heightened_final_step'],
+                                                                      upsample_factor=scale, add_scalable_noise_to_transforms=opt_net['add_noise'])
     elif which_model == 'spsr':
         netG = spsr.SPSRNet(in_nc=opt_net['in_nc'], out_nc=opt_net['out_nc'], nf=opt_net['nf'],
                             nb=opt_net['nb'], upscale=opt_net['scale'])
@@ -160,6 +170,8 @@ def define_D_net(opt_net, img_sz=None, wrap=False):
         netD = SRGAN_arch.CrossCompareDiscriminator(in_nc=opt_net['in_nc'], ref_channels=opt_net['ref_channels'] if 'ref_channels' in opt_net.keys() else 3, nf=opt_net['nf'], scale=opt_net['scale'])
     elif which_model == "discriminator_refvgg":
         netD = SRGAN_arch.RefDiscriminatorVgg128(in_nc=opt_net['in_nc'], nf=opt_net['nf'], input_img_factor=img_sz / 128)
+    elif which_model == "psnr_approximator":
+        netD = SRGAN_arch.PsnrApproximator(nf=opt_net['nf'], input_img_factor=img_sz / 128)
     else:
         raise NotImplementedError('Discriminator model [{:s}] not recognized'.format(which_model))
     return netD

codes/models/steps/injectors.py

@@ -1,3 +1,5 @@
+import random
+
 import torch.nn
 from torch.cuda.amp import autocast
 
@@ -45,6 +47,12 @@ def create_injector(opt_inject, env):
         return ImageFftInjector(opt_inject, env)
     elif type == 'extract_indices':
         return IndicesExtractor(opt_inject, env)
+    elif type == 'random_shift':
+        return RandomShiftInjector(opt_inject, env)
+    elif type == 'psnr':
+        return PsnrInjector(opt_inject, env)
+    elif type == 'batch_rotate':
+        return BatchRotateInjector(opt_inject, env)
     else:
         raise NotImplementedError
 
@@ -94,12 +102,13 @@ class DiscriminatorInjector(Injector):
         super(DiscriminatorInjector, self).__init__(opt, env)
 
     def forward(self, state):
-        d = self.env['discriminators'][self.opt['discriminator']]
+        with autocast(enabled=self.env['opt']['fp16']):
-        if isinstance(self.input, list):
+            d = self.env['discriminators'][self.opt['discriminator']]
-            params = [state[i] for i in self.input]
+            if isinstance(self.input, list):
-            results = d(*params)
+                params = [state[i] for i in self.input]
-        else:
+                results = d(*params)
-            results = d(state[self.input])
+            else:
+                results = d(state[self.input])
         new_state = {}
         if isinstance(self.output, list):
             # Only dereference tuples or lists, not tensors.
@@ -232,10 +241,25 @@ class MarginRemoval(Injector):
     def __init__(self, opt, env):
         super(MarginRemoval, self).__init__(opt, env)
         self.margin = opt['margin']
+        self.random_shift_max = opt['random_shift_max'] if 'random_shift_max' in opt.keys() else 0
 
     def forward(self, state):
         input = state[self.input]
-        return {self.opt['out']: input[:, :, self.margin:-self.margin, self.margin:-self.margin]}
+        if self.random_shift_max > 0:
+            output = []
+            # This is a really shitty way of doing this. If it works at all, I should reconsider using Resample2D, for example.
+            for b in range(input.shape[0]):
+                shiftleft = random.randint(-self.random_shift_max, self.random_shift_max)
+                shifttop = random.randint(-self.random_shift_max, self.random_shift_max)
+                output.append(input[b, :, self.margin+shiftleft:-(self.margin-shiftleft),
+                                 self.margin+shifttop:-(self.margin-shifttop)])
+            output = torch.stack(output, dim=0)
+        else:
+            output = input[:, :, self.margin:-self.margin,
+                                 self.margin:-self.margin]
+
+        return {self.opt['out']: output}
+
 
 # Produces an injection which is composed of applying a single injector multiple times across a single dimension.
 class ForEachInjector(Injector):
@@ -254,7 +278,7 @@ class ForEachInjector(Injector):
         for i in range(inputs.shape[1]):
             st['_in'] = inputs[:, i]
             injs.append(self.injector(st)['_out'])
-        return {self.output: torch.stack(injs, dim=1)}        
+        return {self.output: torch.stack(injs, dim=1)}
 
     
 class ConstantInjector(Injector):
@@ -316,3 +340,31 @@ class IndicesExtractor(Injector):
                 results[o] = state[self.input][:, i]
         return results
 
+
+class RandomShiftInjector(Injector):
+    def __init__(self, opt, env):
+        super(RandomShiftInjector, self).__init__(opt, env)
+
+    def forward(self, state):
+        img = state[self.input]
+        return {self.output: img}
+
+
+class PsnrInjector(Injector):
+    def __init__(self, opt, env):
+        super(PsnrInjector, self).__init__(opt, env)
+
+    def forward(self, state):
+        img1, img2 = state[self.input[0]], state[self.input[1]]
+        mse = torch.mean((img1 - img2) ** 2, dim=[1,2,3])
+        return {self.output: mse}
+
+
+class BatchRotateInjector(Injector):
+    def __init__(self, opt, env):
+        super(BatchRotateInjector, self).__init__(opt, env)
+
+    def forward(self, state):
+        img = state[self.input]
+        return {self.output: torch.roll(img, 1, 0)}
+

codes/process_video.py

@@ -159,8 +159,6 @@ if __name__ == "__main__":
             if 'recurrent_hr_generator' in opt.keys():
                 recurrent_gen = model.env['generators']['generator']
                 model.env['generators']['generator'] = model.env['generators'][opt['recurrent_hr_generator']]
-            else:
-                model.env['generators']['generator'] = recurrent_gen
 
         first_frame = False
         if recurrent_mode:

codes/train2.py

@@ -278,7 +278,7 @@ class Trainer:
 
 if __name__ == '__main__':
     parser = argparse.ArgumentParser()
-    parser.add_argument('-opt', type=str, help='Path to option YAML file.', default='../options/train_exd_mi1_rrdb4x_10bl_bypass.yml')
+    parser.add_argument('-opt', type=str, help='Path to option YAML file.', default='../options/train_exd_mi1_srg2classic_4x.yml')
     parser.add_argument('--launcher', choices=['none', 'pytorch'], default='none', help='job launcher')
     args = parser.parse_args()
     opt = option.parse(args.opt, is_train=True)