Update app.py

app.py

@@ -1,26 +1,61 @@
-
-import gradio as gr
 import torch
-import numpy as np
+import torchvision.transforms as T
 from PIL import Image
-from torchvision import transforms as T
 import joblib
+import json
+import cv2
+import gradio as gr
+
+# Define image transformation
+transform_image = T.Compose([
+    T.ToTensor(),
+    T.Resize(244),
+    T.CenterCrop(224),
+    T.Normalize([0.5], [0.5])
+])
 
-# Load models
-dinov2_vits14 = torch.load('dinov2_vits14.pth', map_location=torch.device('cpu'))
+def load_image(img: str) -> torch.Tensor:
+    """
+    Load an image and return a tensor that can be used as an input to DINOv2.
+    """
+    img = Image.open(img)
+    transformed_img = transform_image(img)[:3].unsqueeze(0)
+    return transformed_img
+
+# Load models for inference
+dinov2_vits14 = torch.hub.load("facebookresearch/dinov2", "dinov2_vits14")
+device = torch.device('cuda' if torch.cuda.is_available() else "cpu")
+dinov2_vits14.to(device)
+dinov2_vits14.eval()  # Set the model to evaluation mode
+
+# Load the classifier
 clf = joblib.load('svm_model.joblib')
 
-# Transform for input image
-transform_image = T.Compose([T.ToTensor(), T.Resize(244), T.CenterCrop(224), T.Normalize([0.5], [0.5])])
+# Load the embeddings
+with open('all_embeddings.json', 'r') as f:
+    embeddings = json.load(f)
 
-def predict(image):
-    image = Image.fromarray(image)
-    transformed_img = transform_image(image)[:3].unsqueeze(0)
+# Predict class for a new image
+def predict_image_class(image_path):
+    new_image = load_image(image_path).to(device)
     with torch.no_grad():
-        embedding = dinov2_vits14(transformed_img)
-        prediction = clf.predict(np.array(embedding[0].cpu()).reshape(1, -1))
+        embedding = dinov2_vits14(new_image).cpu().numpy().reshape(1, -1)
+    prediction = clf.predict(embedding)
     return prediction[0]
 
-iface = gr.Interface(fn=predict, inputs="image", outputs="text")
+# Gradio interface
+def classify_image(image):
+    predicted_class = predict_image_class(image)
+    return f"Predicted class: {predicted_class}"
+
+# Define the Gradio interface
+iface = gr.Interface(
+    fn=classify_image,
+    inputs=gr.Image(type="filepath"),
+    outputs="text",
+    title="Currency Classifier",
+    description="Upload an image of currency to classify."
+)
+
+# Launch the Gradio interface
 iface.launch()
-