Add matcher

matcher/management/commands/populate_products.py

@@ -0,0 +1,55 @@
+
+#
+
+
+import os
+from django.core.management.base import BaseCommand
+from django.conf import settings
+from django.core.files import File
+from matcher.models import Product
+from matcher.similarity import extract_features
+from PIL import Image as PILImage
+
+class Command(BaseCommand):
+    help = "Populate the database with products from images in media/uploads/."
+
+    def handle(self, *args, **kwargs):
+        uploads_dir = os.path.join(settings.MEDIA_ROOT, "uploads")
+
+        if not os.path.exists(uploads_dir):
+            self.stdout.write(self.style.ERROR(f"Folder not found: {uploads_dir}"))
+            return
+
+        # Clear old products
+        Product.objects.all().delete()
+        self.stdout.write("Deleted old products.")
+
+        for filename in os.listdir(uploads_dir):
+            file_path = os.path.join(uploads_dir, filename)
+
+            if not filename.lower().endswith((".jpg", ".jpeg", ".png")):
+                continue
+
+            try:
+                pil_image = PILImage.open(file_path).convert("RGB")
+                features = extract_features(pil_image)
+
+                product = Product(
+                    name=os.path.splitext(filename)[0],
+                    category="General",
+                )
+
+                with open(file_path, "rb") as f:
+                    product.image.save(filename, File(f), save=False)
+
+                if features is not None:
+                    product.feature_vector = features.tobytes()
+
+                product.save()
+                self.stdout.write(self.style.SUCCESS(f"Added {filename}"))
+
+            except Exception as e:
+                self.stdout.write(self.style.ERROR(f"Error processing {filename}: {e}"))
+
+
+        self.stdout.write(self.style.SUCCESS("Finished populating products."))

matcher/matcher/management/commands/populate_products.py

@@ -0,0 +1,55 @@
+
+#
+
+
+import os
+from django.core.management.base import BaseCommand
+from django.conf import settings
+from django.core.files import File
+from matcher.models import Product
+from matcher.similarity import extract_features
+from PIL import Image as PILImage
+
+class Command(BaseCommand):
+    help = "Populate the database with products from images in media/uploads/."
+
+    def handle(self, *args, **kwargs):
+        uploads_dir = os.path.join(settings.MEDIA_ROOT, "uploads")
+
+        if not os.path.exists(uploads_dir):
+            self.stdout.write(self.style.ERROR(f"Folder not found: {uploads_dir}"))
+            return
+
+        # Clear old products
+        Product.objects.all().delete()
+        self.stdout.write("Deleted old products.")
+
+        for filename in os.listdir(uploads_dir):
+            file_path = os.path.join(uploads_dir, filename)
+
+            if not filename.lower().endswith((".jpg", ".jpeg", ".png")):
+                continue
+
+            try:
+                pil_image = PILImage.open(file_path).convert("RGB")
+                features = extract_features(pil_image)
+
+                product = Product(
+                    name=os.path.splitext(filename)[0],
+                    category="General",
+                )
+
+                with open(file_path, "rb") as f:
+                    product.image.save(filename, File(f), save=False)
+
+                if features is not None:
+                    product.feature_vector = features.tobytes()
+
+                product.save()
+                self.stdout.write(self.style.SUCCESS(f"Added {filename}"))
+
+            except Exception as e:
+                self.stdout.write(self.style.ERROR(f"Error processing {filename}: {e}"))
+
+
+        self.stdout.write(self.style.SUCCESS("Finished populating products."))

matcher/matcher/migrations/0001_initial.py

@@ -0,0 +1,32 @@
+# Generated by Django 5.2.5 on 2025-08-26 08:59
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    initial = True
+
+    dependencies = [
+    ]
+
+    operations = [
+        migrations.CreateModel(
+            name='Product',
+            fields=[
+                ('id', models.BigAutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
+                ('name', models.CharField(max_length=255)),
+                ('category', models.CharField(max_length=255)),
+                ('image', models.ImageField(upload_to='products/')),
+                ('feature_vector', models.BinaryField(blank=True, null=True)),
+            ],
+        ),
+        migrations.CreateModel(
+            name='UploadedImage',
+            fields=[
+                ('id', models.BigAutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
+                ('image', models.ImageField(upload_to='uploads/')),
+                ('uploaded_at', models.DateTimeField(auto_now_add=True)),
+            ],
+        ),
+    ]

matcher/matcher/models.py

@@ -0,0 +1,14 @@
+from django.db import models
+
+class Product(models.Model):
+    name = models.CharField(max_length=255)
+    category = models.CharField(max_length=255)
+    image = models.ImageField(upload_to='products/')
+    feature_vector = models.BinaryField(null=True, blank=True)
+
+    def __str__(self):
+        return self.name
+
+class UploadedImage(models.Model):
+    image = models.ImageField(upload_to='uploads/')
+    uploaded_at = models.DateTimeField(auto_now_add=True)

matcher/matcher/similarity.py

@@ -0,0 +1,66 @@
+import torch
+from transformers import ViTImageProcessor, ViTModel
+from PIL import Image
+import numpy as np
+from sklearn.metrics.pairwise import cosine_similarity
+import requests
+from io import BytesIO
+
+# We are now using a much smaller "tiny" version of the Vision Transformer because render was not able to handle and demanding to pay for larger models.
+MODEL_NAME = 'WinKawaks/vit-tiny-patch16-224'
+
+try:
+    image_processor = ViTImageProcessor.from_pretrained(MODEL_NAME)
+    model = ViTModel.from_pretrained(MODEL_NAME)
+except Exception as e:
+    print(f"Error loading model: {e}")
+    image_processor = None
+    model = None
+
+def extract_features(image: Image.Image) -> np.ndarray | None:
+    """
+    Extracts a feature vector from an image using the ViT model.
+    """
+    if not model or not image_processor:
+        print("Model or image processor not loaded.")
+        return None
+        
+    try:
+        inputs = image_processor(images=image, return_tensors="pt")
+
+        with torch.no_grad():
+            outputs = model(**inputs)
+        
+        last_hidden_states = outputs.last_hidden_state
+        features = last_hidden_states.mean(dim=1).cpu().numpy()
+        return features
+    except Exception as e:
+        print(f"An error occurred during feature extraction: {e}")
+        return None
+
+def find_similar_products(uploaded_features, all_products, top_n=20):
+    """
+    Finds similar products by comparing feature vectors using cosine similarity.
+    """
+    if uploaded_features is None or not all_products:
+        return []
+
+    product_features = np.array([np.frombuffer(p.feature_vector, dtype=np.float32) for p in all_products])
+    
+    if product_features.ndim == 1:
+        product_features = product_features.reshape(1, -1)
+    if uploaded_features.ndim == 1:
+        uploaded_features = uploaded_features.reshape(1, -1)
+
+    similarities = cosine_similarity(uploaded_features, product_features)[0]
+
+    similar_products_with_scores = []
+    for i, product in enumerate(all_products):
+        similar_products_with_scores.append({
+            'product': product,
+            'similarity': similarities[i]
+        })
+
+    similar_products_with_scores.sort(key=lambda x: x['similarity'], reverse=True)
+
+    return similar_products_with_scores[:top_n]

matcher/matcher/urls.py

@@ -0,0 +1,6 @@
+from django.urls import path
+from . import views
+
+urlpatterns = [
+    path('', views.index, name='index'),
+]

matcher/matcher/views.py

@@ -0,0 +1,76 @@
+import requests
+from io import BytesIO
+from django.shortcuts import render
+from django.core.files.base import ContentFile
+from .models import Product, UploadedImage
+from .similarity import extract_features, find_similar_products
+from PIL import Image as PILImage
+import numpy as np
+
+def index(request):
+    similar_products = []
+    uploaded_image_url = None
+    error_message = None
+    loading = False
+
+    if request.method == 'POST':
+        loading = True
+        image_file = request.FILES.get('image_file')
+        image_url_from_post = request.POST.get('image_url')
+        
+        image_to_process = None
+        image_bytes = None
+        image_filename = 'uploaded_image.jpg' 
+
+        try:
+            if image_file:
+                image_filename = image_file.name
+                image_bytes = image_file.read()
+                image_to_process = PILImage.open(BytesIO(image_bytes)).convert("RGB")
+
+            elif image_url_from_post:
+                headers = {'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'}
+                response = requests.get(image_url_from_post, headers=headers, timeout=15)
+                response.raise_for_status()
+                image_bytes = response.content
+                image_to_process = PILImage.open(BytesIO(image_bytes)).convert("RGB")
+                image_filename = image_url_from_post.split('/')[-1]
+
+            else:
+                error_message = "Please upload an image or provide a URL."
+
+            if image_to_process and image_bytes:
+                uploaded_image_instance = UploadedImage()
+                uploaded_image_instance.image.save(image_filename, ContentFile(image_bytes))
+                
+                uploaded_image_url = uploaded_image_instance.image.url
+                
+                uploaded_features = extract_features(image_to_process)
+
+                if uploaded_features is not None:
+                    all_products = list(Product.objects.exclude(feature_vector__isnull=True))
+                    similar_products = find_similar_products(uploaded_features, all_products)
+
+                    min_similarity = request.POST.get('similarity_score')
+                    if min_similarity:
+                        min_similarity = float(min_similarity)
+                        similar_products = [p for p in similar_products if p['similarity'] >= min_similarity]
+                else:
+                    error_message = "Could not extract features from the image."
+
+        except PILImage.UnidentifiedImageError:
+            error_message = "Could not identify the file as an image. Please check the file or URL."
+        except requests.exceptions.RequestException as e:
+            error_message = f"Failed to retrieve image from URL: {e}"
+        except Exception as e:
+            error_message = f"An unexpected error occurred: {e}"
+        
+        loading = False
+
+    return render(request, 'matcher/index.html', {
+        'similar_products': similar_products,
+        'uploaded_image_url': uploaded_image_url,
+        'error_message': error_message,
+        'loading': loading,
+    })
+

matcher/migrations/0001_initial.py

@@ -0,0 +1,32 @@
+# Generated by Django 5.2.5 on 2025-08-26 08:59
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    initial = True
+
+    dependencies = [
+    ]
+
+    operations = [
+        migrations.CreateModel(
+            name='Product',
+            fields=[
+                ('id', models.BigAutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
+                ('name', models.CharField(max_length=255)),
+                ('category', models.CharField(max_length=255)),
+                ('image', models.ImageField(upload_to='products/')),
+                ('feature_vector', models.BinaryField(blank=True, null=True)),
+            ],
+        ),
+        migrations.CreateModel(
+            name='UploadedImage',
+            fields=[
+                ('id', models.BigAutoField(auto_created=True, primary_key=True, serialize=False, verbose_name='ID')),
+                ('image', models.ImageField(upload_to='uploads/')),
+                ('uploaded_at', models.DateTimeField(auto_now_add=True)),
+            ],
+        ),
+    ]

matcher/models.py

@@ -0,0 +1,14 @@
+from django.db import models
+
+class Product(models.Model):
+    name = models.CharField(max_length=255)
+    category = models.CharField(max_length=255)
+    image = models.ImageField(upload_to='products/')
+    feature_vector = models.BinaryField(null=True, blank=True)
+
+    def __str__(self):
+        return self.name
+
+class UploadedImage(models.Model):
+    image = models.ImageField(upload_to='uploads/')
+    uploaded_at = models.DateTimeField(auto_now_add=True)

matcher/similarity.py

@@ -0,0 +1,66 @@
+import torch
+from transformers import ViTImageProcessor, ViTModel
+from PIL import Image
+import numpy as np
+from sklearn.metrics.pairwise import cosine_similarity
+import requests
+from io import BytesIO
+
+# We are now using a much smaller "tiny" version of the Vision Transformer because render was not able to handle and demanding to pay for larger models.
+MODEL_NAME = 'WinKawaks/vit-tiny-patch16-224'
+
+try:
+    image_processor = ViTImageProcessor.from_pretrained(MODEL_NAME)
+    model = ViTModel.from_pretrained(MODEL_NAME)
+except Exception as e:
+    print(f"Error loading model: {e}")
+    image_processor = None
+    model = None
+
+def extract_features(image: Image.Image) -> np.ndarray | None:
+    """
+    Extracts a feature vector from an image using the ViT model.
+    """
+    if not model or not image_processor:
+        print("Model or image processor not loaded.")
+        return None
+        
+    try:
+        inputs = image_processor(images=image, return_tensors="pt")
+
+        with torch.no_grad():
+            outputs = model(**inputs)
+        
+        last_hidden_states = outputs.last_hidden_state
+        features = last_hidden_states.mean(dim=1).cpu().numpy()
+        return features
+    except Exception as e:
+        print(f"An error occurred during feature extraction: {e}")
+        return None
+
+def find_similar_products(uploaded_features, all_products, top_n=20):
+    """
+    Finds similar products by comparing feature vectors using cosine similarity.
+    """
+    if uploaded_features is None or not all_products:
+        return []
+
+    product_features = np.array([np.frombuffer(p.feature_vector, dtype=np.float32) for p in all_products])
+    
+    if product_features.ndim == 1:
+        product_features = product_features.reshape(1, -1)
+    if uploaded_features.ndim == 1:
+        uploaded_features = uploaded_features.reshape(1, -1)
+
+    similarities = cosine_similarity(uploaded_features, product_features)[0]
+
+    similar_products_with_scores = []
+    for i, product in enumerate(all_products):
+        similar_products_with_scores.append({
+            'product': product,
+            'similarity': similarities[i]
+        })
+
+    similar_products_with_scores.sort(key=lambda x: x['similarity'], reverse=True)
+
+    return similar_products_with_scores[:top_n]

matcher/urls.py

@@ -0,0 +1,6 @@
+from django.urls import path
+from . import views
+
+urlpatterns = [
+    path('', views.index, name='index'),
+]

matcher/views.py

@@ -0,0 +1,76 @@
+import requests
+from io import BytesIO
+from django.shortcuts import render
+from django.core.files.base import ContentFile
+from .models import Product, UploadedImage
+from .similarity import extract_features, find_similar_products
+from PIL import Image as PILImage
+import numpy as np
+
+def index(request):
+    similar_products = []
+    uploaded_image_url = None
+    error_message = None
+    loading = False
+
+    if request.method == 'POST':
+        loading = True
+        image_file = request.FILES.get('image_file')
+        image_url_from_post = request.POST.get('image_url')
+        
+        image_to_process = None
+        image_bytes = None
+        image_filename = 'uploaded_image.jpg' 
+
+        try:
+            if image_file:
+                image_filename = image_file.name
+                image_bytes = image_file.read()
+                image_to_process = PILImage.open(BytesIO(image_bytes)).convert("RGB")
+
+            elif image_url_from_post:
+                headers = {'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'}
+                response = requests.get(image_url_from_post, headers=headers, timeout=15)
+                response.raise_for_status()
+                image_bytes = response.content
+                image_to_process = PILImage.open(BytesIO(image_bytes)).convert("RGB")
+                image_filename = image_url_from_post.split('/')[-1]
+
+            else:
+                error_message = "Please upload an image or provide a URL."
+
+            if image_to_process and image_bytes:
+                uploaded_image_instance = UploadedImage()
+                uploaded_image_instance.image.save(image_filename, ContentFile(image_bytes))
+                
+                uploaded_image_url = uploaded_image_instance.image.url
+                
+                uploaded_features = extract_features(image_to_process)
+
+                if uploaded_features is not None:
+                    all_products = list(Product.objects.exclude(feature_vector__isnull=True))
+                    similar_products = find_similar_products(uploaded_features, all_products)
+
+                    min_similarity = request.POST.get('similarity_score')
+                    if min_similarity:
+                        min_similarity = float(min_similarity)
+                        similar_products = [p for p in similar_products if p['similarity'] >= min_similarity]
+                else:
+                    error_message = "Could not extract features from the image."
+
+        except PILImage.UnidentifiedImageError:
+            error_message = "Could not identify the file as an image. Please check the file or URL."
+        except requests.exceptions.RequestException as e:
+            error_message = f"Failed to retrieve image from URL: {e}"
+        except Exception as e:
+            error_message = f"An unexpected error occurred: {e}"
+        
+        loading = False
+
+    return render(request, 'matcher/index.html', {
+        'similar_products': similar_products,
+        'uploaded_image_url': uploaded_image_url,
+        'error_message': error_message,
+        'loading': loading,
+    })
+