Update src/streamlit_app.py

src/streamlit_app.py

@@ -56,7 +56,7 @@ def show_img(img_id, score=None, col=None):
 
                 text = '- '.join(parts)
                 if score:
-                    text += f'\n Score: {score}'
+                    text += f'\n\n Score: {score:.2f}'
 
                 # Hiển thị ảnh trong cột
                 if col:
@@ -84,8 +84,8 @@ def search_faiss(model, processor, index, id_map, prompt, top_k=5, device='cpu')
     D, I = index.search(q, top_k)
 
     st.write("FAISS search completed.")
-    st.write("Indices returned:", I[0])
-    st.write("Scores returned:", D[0])
+    # st.write("Indices returned:", I[0])
+    # st.write("Scores returned:", D[0])
     # st.write("ID map keys sample:", list(id_map.keys())[:10])
     return [(id_map[i], float(D[0][j])) for j, i in enumerate(I[0])]
 
@@ -102,7 +102,7 @@ def running(prompt, top_k=5):
     cols = st.columns(5)  # Chia thành 5 cột
     col_idx = 0
     for img_id, score in results:
-        st.write(f"results: {img_id} và {score}")
+        # st.write(f"results: {img_id} và {score}")
         show_img(img_id, score, col=cols[col_idx])
         col_idx += 1
         if col_idx == 5:  # Sau khi hiển thị 5 ảnh, reset cột
@@ -165,12 +165,12 @@ processor = CLIPProcessor.from_pretrained(
 with open(mapping_path, "rb") as f:
     id_map = pickle.load(f)
 
-st.title("Fashion Product Image-Text Retrieval")
+st.title("Fashion Product Image Retrieval")
 
 st.markdown("""
 ### **Overview**
 
-In this project, I demonstrate an **Image-Text Retrieval** system for fashion products. The system uses a fine-tuned **CLIP model** (`clip-vit-base-patch32`) to match images with relevant text descriptions. We have a dataset of **1000 fashion product images**, stored on **Amazon S3**. Each image is associated with detailed product descriptions, such as **product type**, **color**, **category**, and **brand**.
+In this project, I demonstrate an **Image Retrieval** system for fashion products. The system uses a fine-tuned **CLIP model** (`clip-vit-base-patch32`) to match images with relevant text descriptions. We have a dataset of **1000 fashion product images**, stored on **Amazon S3**. Each image is associated with detailed product descriptions, such as **product type**, **color**, **category**, and **brand**.
 
 The goal of this system is to retrieve the most relevant fashion images based on a given text prompt (e.g., "red dress") and vice versa. With this system, users can search for fashion products in a more intuitive, text-based manner.
 
@@ -182,14 +182,14 @@ The goal of this system is to retrieve the most relevant fashion images based on
 """)
 
 # Example to show some images
-st.subheader("Some Sample Images and Their Captions:")
+st.subheader("Some sample images and their captions:")
 example = [13422, 10037, 38246, 23273, 2008]
 example_cols = st.columns(5)  # Chia thành 5 cột
 for idx, img_id in enumerate(example):
     show_img(img_id, None, example_cols[idx])
 
 # Chạy ví dụ với prompt
-st.subheader("Example Usage: Enter a Prompt to Retrieve Related Images")
+st.subheader("Example usage: enter a prompt to retrieve related images")
 prompt_input = st.text_input("Enter a prompt", "a red Apparel dress")
 top_k_input = st.number_input("Enter the number of results (top_k)", min_value=1, max_value=10, value=5)