Add application file

app.py

@@ -0,0 +1,46 @@
+import torch
+import torch.nn as nn
+from PIL import Image
+from torchvision import transforms
+import os
+import gradio as gr
+device = torch.device("mps")
+class UPSC(nn.Module):
+    def __init__(self):
+        super(UPSC,self).__init__()
+        self.model = nn.Sequential(
+            nn.Conv2d(in_channels=3, out_channels=64, kernel_size=5, padding=2),
+            nn.ReLU(),
+            nn.Conv2d(in_channels=64, out_channels=32, kernel_size=3, padding=1),
+            nn.ReLU(),
+            # This convolution outputs channels that are scale_factor^2 * number_of_channels.
+            nn.Conv2d(in_channels=32, out_channels=3 * 3 * 3, kernel_size=3, padding=1),
+            # PixelShuffle rearranges channels into spatial dimensions.
+            nn.PixelShuffle(3)
+        )
+    def forward(self, x):
+        return self.model(x)
+
+model = UPSC().to(device)
+model.load_state_dict(torch.load("upscaling.pth", weights_only=True))
+model.eval()
+
+def task(img):
+    transform = transforms.Compose([
+        transforms.Resize((256, 256)),  # match training input size
+        transforms.ToTensor()
+    ])
+    
+    lr_tensor = transform(img).unsqueeze(0).to(device)
+    
+    with torch.no_grad():
+        sr_tensor = model(lr_tensor)
+    
+    # Remove batch dimension and convert to PIL
+    sr_image = transforms.ToPILImage()(sr_tensor.squeeze(0).clamp(0, 1))
+    return sr_image
+gr.Interface(fn=task,
+             inputs=gr.Image(type="pil"),
+             outputs=gr.Image(type="pil"),
+             title="Image Upscaling",
+             description="Upload an image to upscale it using a neural network.").launch(share=True)

upscaling.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8b27d159a451b1fac7efc1d1e3b2828dfafeea2695d344249df6a4cbf312f1b
+size 127260