Image Processing with Pytorch

ImageProcessingTorch/im_calc.py

@@ -0,0 +1,65 @@
+import torch
+import torch.nn.functional as F
+from torchvision.transforms import ToTensor
+from PIL import Image
+from torchvision.models import vgg16
+import torchvision.transforms as transforms
+
+# Upoading reference and target image
+reference_image = Image.open('ref1-0.jpg').convert('RGB')
+target_image = Image.open('tar2-0.jpg').convert('RGB')
+
+# Converting images to Tensor format
+reference_tensor = ToTensor()(reference_image)
+target_tensor = ToTensor()(target_image)
+
+# Check the size of the images and resize them if necessary
+if reference_tensor.dim() != 3 or target_tensor.dim() != 3:
+    print("Image sizes are mismatched. Use RGB images.")
+    exit()
+
+if reference_tensor.shape != target_tensor.shape:
+    print("Reference and target images must be equal in size.")
+    exit()
+
+# Reshape dimensions
+reference_tensor = reference_tensor.unsqueeze(0)  # Add batch size
+target_tensor = target_tensor.unsqueeze(0)
+
+# Moving image tensors to the GPU
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+reference_tensor = reference_tensor.to(device)
+target_tensor = target_tensor.to(device)
+
+# Loading the VGG16 model
+model = vgg16(pretrained=True)
+model = model.to(device)
+
+# Processing images appropriately for the VGG16 model
+preprocess = transforms.Compose([
+    transforms.Resize((224, 224)),  # Resize images to VGG16 input size
+    transforms.ToTensor(),  # Converting tensor format
+    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])  # Normalizing
+])
+
+reference_tensor = preprocess(reference_image).unsqueeze(0).to(device)
+target_tensor = preprocess(target_image).unsqueeze(0).to(device)
+
+# Similarity Calculating Function
+def calculate_similarity(reference_tensor, target_tensor):
+    # Feature maps of reference and target images
+    reference_features = model.features(reference_tensor)
+    target_features = model.features(target_tensor)
+
+    # Similarity Calculating
+    similarity = F.cosine_similarity(reference_features, target_features).mean()
+
+    return similarity
+
+# Calculating of Similarity
+similarity_score = calculate_similarity(reference_tensor, target_tensor)
+
+# Similarity percenatage
+similarity_percentage = similarity_score.item() * 100
+
+print(f"Benzerlik Skoru: {similarity_percentage:.2f}%")