Update src/streamlit_app.py

src/streamlit_app.py

@@ -1,4 +1,5 @@
 import streamlit as st
+from huggingface_hub import InferenceClient
 import requests
 from PIL import Image
 import tempfile
@@ -16,23 +17,18 @@ st.set_page_config(page_title="Cassava Disease Detection", layout="centered")
 # -----------------------------
 ROBOFLOW_API_KEY = st.secrets["ROBOFLOW_API_KEY"]
 OPENROUTER_API_KEY = st.secrets["OPENROUTER_API_KEY"]
-
 MODEL_ID = "cassavadisease/1"
-ROBOFLOW_API_URL = "https://serverless.roboflow.com"
+ROBOFLOW_API_URL = "https://detect.roboflow.com"
 
 # -----------------------------
-# INITIALIZE ROBOFLOW CLIENT
+# INITIALIZE HF INFERENCE CLIENT
 # -----------------------------
-CLIENT = InferenceHTTPClient(
-    api_url=ROBOFLOW_API_URL,
-    api_key=ROBOFLOW_API_KEY
-)
+hf_client = InferenceClient(token=OPENROUTER_API_KEY)
 
 # -----------------------------
 # FUNCTION: AI EXPLANATION
 # -----------------------------
 def get_ai_explanation(disease_name):
-
     prompt = f"""
     Explain briefly the cassava disease: {disease_name}.
     Include:
@@ -42,35 +38,8 @@ def get_ai_explanation(disease_name):
     - Treatment
     Keep answer short.
     """
-
-    response = requests.post(
-        "https://openrouter.ai/api/v1/chat/completions",
-        headers={
-            "Authorization": f"Bearer {OPENROUTER_API_KEY}",
-            "Content-Type": "application/json",
-            "HTTP-Referer": "http://localhost:8501",
-            "X-Title": "Cassava Disease Detection App"
-        },
-        json={
-            "model": "minimax/minimax-m2.5",
-            "messages": [
-                {"role": "user", "content": prompt}
-            ],
-            "max_tokens": 800,
-            "temperature": 0.3
-        }
-    )
-
-    if response.status_code != 200:
-        return f"OpenRouter API Error:\n{response.text}"
-
-    result = response.json()
-
-    if "choices" not in result:
-        return f"Unexpected API Response:\n{result}"
-
-    return result["choices"][0]["message"]["content"]
-
+    response = hf_client.chat(model="minimax/minimax-m2.5", inputs=prompt)
+    return response.get("generated_text") if "generated_text" in response else str(response)
 
 # -----------------------------
 # UI
@@ -79,14 +48,12 @@ st.title("Cassava Disease Detection Web App")
 st.write("Upload or capture a cassava leaf image for disease detection.")
 
 source = st.radio("Select Image Source:", ["Upload Image", "Use Camera"])
-
 image = None
 
 if source == "Upload Image":
     uploaded_file = st.file_uploader("Upload Image", type=["jpg", "jpeg", "png"])
     if uploaded_file:
         image = Image.open(uploaded_file)
-
 elif source == "Use Camera":
     camera_photo = st.camera_input("Take a picture of the cassava leaf")
     if camera_photo:
@@ -96,7 +63,6 @@ elif source == "Use Camera":
 # MAIN PROCESS
 # -----------------------------
 if image is not None:
-
     st.image(image, caption="Captured Image", use_container_width=True)
 
     # Save temp image
@@ -104,56 +70,40 @@ if image is not None:
         image.save(tmp.name)
         temp_path = tmp.name
 
-    # Roboflow inference
+    # -----------------------------
+    # Roboflow inference (via requests)
+    # -----------------------------
     with st.spinner("Analyzing image..."):
-        result = CLIENT.infer(temp_path, model_id=MODEL_ID)
-
+        headers = {"Authorization": f"Bearer {ROBOFLOW_API_KEY}"}
+        files = {"file": open(temp_path, "rb")}
+        response = requests.post(f"{ROBOFLOW_API_URL}/{MODEL_ID}", headers=headers, files=files)
+        result = response.json()
+    
     os.remove(temp_path)
-
+    
     predictions = result.get("predictions", [])
-
+    
     if predictions:
-
         img_cv = np.array(image)
         img_cv = cv2.cvtColor(img_cv, cv2.COLOR_RGB2BGR)
 
         for pred in predictions:
-
-            x = pred["x"]
-            y = pred["y"]
-            w = pred["width"]
-            h = pred["height"]
+            x, y, w, h = pred["x"], pred["y"], pred["width"], pred["height"]
             label = pred["class"]
             confidence = round(pred["confidence"] * 100, 2)
 
-            # Convert center to corner format
             x1 = int(x - w / 2)
             y1 = int(y - h / 2)
             x2 = int(x + w / 2)
             y2 = int(y + h / 2)
 
-            # Draw bounding box
             cv2.rectangle(img_cv, (x1, y1), (x2, y2), (0, 255, 0), 2)
-
-            # Label background
             cv2.rectangle(img_cv, (x1, y1 - 30), (x1 + 250, y1), (0, 255, 0), -1)
+            cv2.putText(img_cv, f"{label} ({confidence}%)", (x1 + 5, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0,0,0), 2)
 
-            # Label text
-            cv2.putText(
-                img_cv,
-                f"{label} ({confidence}%)",
-                (x1 + 5, y1 - 10),
-                cv2.FONT_HERSHEY_SIMPLEX,
-                0.6,
-                (0, 0, 0),
-                2
-            )
-
-        # Convert back to RGB
         img_display = cv2.cvtColor(img_cv, cv2.COLOR_BGR2RGB)
         st.image(img_display, caption="Detected & Labeled Image", use_container_width=True)
 
-        # Get highest confidence prediction
         top_prediction = max(predictions, key=lambda x: x["confidence"])
         disease_name = top_prediction["class"]
         confidence = round(top_prediction["confidence"] * 100, 2)
@@ -161,17 +111,13 @@ if image is not None:
         st.success(f"Detected: **{disease_name}**")
         st.info(f"Confidence: {confidence}%")
 
-        # AI Explanation
         with st.spinner("Generating disease explanation..."):
             explanation = get_ai_explanation(disease_name)
-
         st.markdown("## 📘 Disease Information")
         st.write(explanation)
-
     else:
         st.warning("No cassava leaf detected.")
 
-
 # -----------------------------
 # FOOTER
 # -----------------------------