Update app.py (#44)

- Update app.py (5d53f733bed47a5e0f1b900a9e3bd7ea639ec15b)

app.py

@@ -14,7 +14,6 @@ import requests
 GROQ_API_KEY = "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"
 GROQ_MODEL = "llama3-70b-8192"
 
-# Load BLIP and SentenceTransformer
 @st.cache_resource(show_spinner=False)
 def load_models():
     processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
@@ -24,7 +23,6 @@ def load_models():
 
 processor, blip_model, embedder = load_models()
 
-# Frame Extraction
 def extract_frames(video_path, interval_sec=1):
     cap = cv2.VideoCapture(video_path)
     fps = cap.get(cv2.CAP_PROP_FPS) or 30
@@ -41,7 +39,6 @@ def extract_frames(video_path, interval_sec=1):
     cap.release()
     return frames
 
-# Caption using BLIP
 def caption_frame(image, processor, model):
     device = "cuda" if torch.cuda.is_available() else "cpu"
     model.to(device)
@@ -50,7 +47,6 @@ def caption_frame(image, processor, model):
     caption = processor.decode(out[0], skip_special_tokens=True)
     return caption
 
-# Call Groq API
 def call_groq_api(prompt):
     url = "https://api.groq.com/openai/v1/chat/completions"
     headers = {
@@ -60,7 +56,7 @@ def call_groq_api(prompt):
     data = {
         "model": GROQ_MODEL,
         "messages": [
-            {"role": "system", "content": "You are an expert cricket video summarizer and analyst."},
+            {"role": "system", "content": "You are an expert video summarizer and scene analyzer."},
             {"role": "user", "content": prompt}
         ],
         "temperature": 0.7,
@@ -69,7 +65,6 @@ def call_groq_api(prompt):
     response.raise_for_status()
     return response.json()["choices"][0]["message"]["content"]
 
-# Create FAISS index
 def create_faiss_index(texts):
     embeddings = embedder.encode(texts, convert_to_numpy=True)
     dim = embeddings.shape[1]
@@ -77,30 +72,26 @@ def create_faiss_index(texts):
     index.add(embeddings)
     return index, embeddings
 
-# Search RAG facts
 def rag_search(query, index, texts, embeddings, top_k=3):
     query_vec = embedder.encode([query], convert_to_numpy=True)
     D, I = index.search(query_vec, top_k)
     return [texts[i] for i in I[0]]
 
-# Knowledge base
 knowledge_base = [
-    "A six is when a batsman hits the ball over the boundary without it touching the ground.",
-    "An out occurs when a batsman is dismissed by the fielding team.",
-    "The umpire raises a finger to signal an out.",
-    "When a player scores a century, the crowd often stands and applauds.",
-    "A bowler appeals by shouting if they believe a batsman is out.",
-    "The third umpire is used for reviewing close decisions.",
-    "Fielders celebrate when a wicket is taken.",
-    "A four is when the ball reaches the boundary after bouncing at least once."
+    "People cheer when something exciting happens in a video.",
+    "A person scoring or celebrating indicates a major event.",
+    "When people run or point, it's likely something important occurred.",
+    "Spectators usually react to notable plays or events.",
+    "Gestures and facial expressions convey emotions in a scene.",
+    "A presenter with a mic is likely commenting on a performance.",
+    "Slow motion replays are usually shown for critical events."
 ]
 
 index, embeddings = create_faiss_index(knowledge_base)
 
-# Streamlit App
-st.title("🏏 Cricket Video Analyzer with RAG, BLIP & Groq")
+st.title("🎥 General Video Analyzer with RAG, BLIP & Groq")
 
-uploaded_file = st.file_uploader("📁 Upload a video file", type=["mp4", "mov", "avi"])
+uploaded_file = st.file_uploader("📁 Upload any video file", type=["mp4", "mov", "avi"])
 
 if uploaded_file is not None:
     with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as tmp:
@@ -120,27 +111,27 @@ if uploaded_file is not None:
     for i, frame in enumerate(frames):
         caption = caption_frame(frame, processor, blip_model)
         captions.append(f"Scene {i+1}: {caption}")
-        st.image(frame, caption=captions[-1], use_column_width=True)
+        st.image(frame, caption=captions[-1], use_container_width=True)
 
     combined_description = " ".join(captions)
 
-    st.info("🔍 Performing RAG search for factual insights...")
+    st.info("🔍 Performing RAG search for contextual facts...")
     facts = rag_search(combined_description, index, knowledge_base, embeddings)
 
-    st.subheader("📌 Retrieved Relevant Cricket Facts:")
+    st.subheader("📌 Related Insights:")
     for fact in facts:
         st.markdown(f"- {fact}")
 
-    st.info("🤖 Generating summary using Groq's LLaMA 3...")
+    st.info("🤖 Generating detailed summary using Groq...")
     prompt = (
-        "Analyze the following cricket scenes and give a detailed summary:\n\n"
+        "Here are descriptions of scenes from a video:\n\n"
         + "\n".join(captions)
-        + "\n\nAlso incorporate cricket facts if relevant."
+        + "\n\nGive a structured summary mentioning key events, people involved, and possible activities."
     )
 
     try:
         summary = call_groq_api(prompt)
-        st.subheader("📝 AI Summary of Video:")
+        st.subheader("📝 Video Summary:")
         st.write(summary)
     except Exception as e:
         st.error(f"Groq API Error: {e}")