PixelPeep

requirements.txt

@@ -0,0 +1,5 @@
+torch
+torchvision
+numpy
+Pillow
+scikit-learn

test_images.py

@@ -0,0 +1,25 @@
+from sklearn.metrics.pairwise import cosine_similarity
+from train import originals, names, get_embedding
+import numpy as np
+
+def test_images(test_path,originals,names,threshold=0.88):
+    #originals, names = load_original_embeddings(original_folder)
+    test_emb = get_embedding(test_path)
+    sims = cosine_similarity(test_emb, originals)[0]
+    best_idx = np.argmax(sims)
+    best_score = sims[best_idx]
+
+    print(f"Best match: {names[best_idx]}")
+    print(f"Similarity score: {best_score:.3f}")
+
+    if best_score >= threshold:
+        print(" Match found !")
+    else:
+        print(" No match found.")
+
+test_images(
+    test_path=r"C:\\Users\\LENOVO\\Documents\\PROJECTS\\pixel-peep\\images\\test\\meme.png",
+    originals=originals,
+    names=names,
+    threshold=0.80
+)

train.py

@@ -0,0 +1,66 @@
+import os
+import torch
+import numpy as np
+from PIL import Image
+from torchvision import models, transforms
+#from sklearn.metrics.pairwise import cosine_similarity
+
+
+resnet = models.resnet50(pretrained=True)
+resnet = torch.nn.Sequential(*list(resnet.children())[:-1])  # Remove layer because it is not needed (last layers) only needs features//classifier
+resnet.eval()
+
+
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor(),
+
+])
+
+def get_embedding(image_path):
+    image = Image.open(image_path).convert("RGB")
+    image = transform(image).unsqueeze(0)
+    with torch.no_grad():
+        embedding = resnet(image).squeeze().numpy()
+    return embedding.reshape(1, -1)
+
+
+def train_images(folder_path):
+    embeddings = []
+    filenames = []
+    for file in os.listdir(folder_path):
+        if file.lower().endswith((".jpg", ".png")):
+            path = os.path.join(folder_path, file)
+            emb = get_embedding(path)
+            embeddings.append(emb)
+            filenames.append(file)
+    return np.vstack(embeddings), filenames
+
+original_folder=r"C:\\Users\\LENOVO\\Documents\\PROJECTS\\pixel-peep\\images\\originals"
+originals, names = train_images(original_folder)
+
+#never mind this part 
+'''
+# Match test image to originals
+def test_images(test_path,originals,names,threshold=0.88):
+    #originals, names = load_original_embeddings(original_folder)
+    test_emb = get_embedding(test_path)
+    sims = cosine_similarity(test_emb, originals)[0]
+    best_idx = np.argmax(sims)
+    best_score = sims[best_idx]
+
+    print(f"Best match: {names[best_idx]}")
+    print(f"Similarity score: {best_score:.3f}")
+
+    if best_score >= threshold:
+        print(" Match found !")
+    else:
+        print(" No match found.")
+#original_folder=r"C:\\Users\\LENOVO\\Documents\\PROJECTS\\pixel-peep\\images\\originals"
+#originals, names = load_original_embeddings(original_folder)
+#test_images(
+    test_path=r"C:\\Users\\LENOVO\\Documents\\PROJECTS\\pixel-peep\\images\\test\\meme.png",
+    originals=originals,
+    names=names,
+    threshold=0.88
+#)'''