Update app.py

app.py

@@ -32,7 +32,7 @@ def similarity_all(query_image_name, embeddings, metric):
     scores = {image_name : metric(querry_embedding, embeddings[image_name]) for image_name in images}
     return scores
 
-def l1_similarity_score(query_embedding, target_embedding):    
+def euclidean_similarity_score(query_embedding, target_embedding):    
     return np.linalg.norm(query_embedding-target_embedding)
 
 def cosine_similarity_score(query_embedding, target_embedding):
@@ -47,8 +47,8 @@ def retrieve(query_image_name, embeddings_type, metric_type):
     else :
         embeddings = naive_embeddings
 
-    if metric_type == 'L1 Norm' :
-        metric = l1_similarity_score
+    if metric_type == 'Euclidean' :
+        metric = euclidean_similarity_score
     else :
         metric = cosine_similarity_score
     
@@ -59,7 +59,7 @@ def retrieve(query_image_name, embeddings_type, metric_type):
 
 input_button = gr.inputs.Dropdown(query_images, label='Choice of the query image')
 embeddings_selection = gr.inputs.Radio(['MobileNetV2', 'BoVW', 'Baseline'], label='Embeddings')
-metric_selection = gr.inputs.Radio(['L1 Norm', 'Cosine'], label='Similarity Metric')
+metric_selection = gr.inputs.Radio(['Euclidean', 'Cosine'], label='Similarity Metric')
 retrieved_images = gr.outputs.Carousel(["image"]*10, label='Retrieved images')
 
 description = "This is a demo of the content-based image retrieval system developed as part of the IR course project, 2022. The indexed dataset is [INRIA Holidays](https://lear.inrialpes.fr/~jegou/data.php). \n\nSeveral image embeddings can be used :\n \n-**MobileNetV2** : feature extraction is performed using a MobileNet architecture trained on ImageNet.\n\n-**BoVW (Bag of Visual Words)** : embedding is the BoVW histogram using color histogram as a descriptor.\n\n-**Baseline** : basic descriptor that uses pixel values of the downsized images."