project push

.gitignore

@@ -0,0 +1,3 @@
+data
+src/__pycache__*
+__pycache__*

app.py

@@ -0,0 +1,38 @@
+import streamlit as st
+from src.multimodelsearch import MultiModelSearch
+st.set_page_config(
+    layout="wide",
+    page_title="Recommendation_engine"
+)
+
+
+def main():
+    st.markdown("<h1 style = 'text-align:center; color:black;'>Recommendation_engine</h1>",unsafe_allow_html=True)
+    
+    multimodelserch = MultiModelSearch()
+
+    query = st.text_input("Enter Your Query")
+
+    if st.button("Search") and len(query) > 0:
+        result = multimodelserch.search(query=query)
+        st.info(f"Your query : {query}")
+        st.subheader("Search Results")
+        col1,col2,col3 = st.columns(3)
+
+        with col1:
+            st.write(f"Score: {round(result[0].score*100)}%")
+            st.image(result[0].content,use_column_width=True)
+    
+        with col2:
+            st.write(f"Score: {round(result[1].score*100)}%")
+            st.image(result[1].content,use_column_width=True)
+
+        with col3:
+            st.write(f"Score: {round(result[2].score*100)}%")
+            st.image(result[2].content,use_column_width=True)
+
+    else:
+        st.warning("Please Enter query.......")
+
+if __name__ == "__main__":
+    main()

config.py

@@ -0,0 +1,2 @@
+MODEL_DIM = 512
+MODEL_NAME = 'sentence-transformers/clip-ViT-B-32'

requirements.txt

@@ -0,0 +1,5 @@
+sentence-transformers
+streamlit
+torch
+farm-haystack
+

src/multimodelsearch.py

@@ -0,0 +1,33 @@
+import os
+from haystack import Document
+from haystack import Pipeline
+from haystack.document_stores import InMemoryDocumentStore
+from haystack.nodes.retriever.multimodal import MultiModalRetriever
+from config import MODEL_DIM, MODEL_NAME
+
+class MultiModelSearch:
+    def __init__(self):
+        self.document_stores = InMemoryDocumentStore(embedding_dim=MODEL_DIM)
+        document_directory = os.path.join(os.getcwd(),"data")
+        # fetch all images and write into haystack document
+        images = [
+            Document(content= f"{document_directory}/{filename}",content_type="image" )
+            for filename in os.listdir(document_directory)
+        ]
+
+        self.document_stores.write_documents(images)
+        self.retriever_text_to_image = MultiModalRetriever(
+            document_store= self.document_stores,
+            query_embedding_model= MODEL_NAME,
+            query_type="text",
+            document_embedding_models= {"image":MODEL_NAME},
+)
+        
+        self.document_stores.update_embeddings(retriever=self.retriever_text_to_image)
+
+        self.pipeline = Pipeline()
+        self.pipeline.add_node(component=self.retriever_text_to_image, name="retriever_text_to_image", inputs=["Query"])
+
+    def search(self,query, top_k = 3):
+        results = self.pipeline.run(query=query, params={"retriever_text_to_image": {"top_k":top_k}})
+        return sorted(results["documents"],key= lambda d:d.score, reverse=True)