a1

app.py

@@ -1,3 +1,7 @@
+import os
+# Force the use of legacy Keras (Keras 2 behavior) so that hub.KerasLayer is recognized properly.
+os.environ["TF_USE_LEGACY_KERAS"] = "1"
+
 import streamlit as st
 import tensorflow as tf
 import tensorflow_hub as hub
@@ -7,7 +11,6 @@ import requests
 from googletrans import Translator
 import asyncio
 import nest_asyncio
-import os
 
 # Allow nested event loops in Streamlit
 nest_asyncio.apply()
@@ -19,7 +22,7 @@ st.set_page_config(
     initial_sidebar_state="expanded",
 )
 
-# Custom CSS styling
+# Custom CSS for styling
 custom_css = """
 <style>
 body { background-color: #f8f9fa; }
@@ -30,6 +33,7 @@ h1, h2, h3, h4 { color: #2c3e50; font-family: 'Segoe UI', Tahoma, Geneva, Verdan
 """
 st.markdown(custom_css, unsafe_allow_html=True)
 
+# Dictionary mapping diseases to recommended pesticides
 pesticide_recommendations = {
     'Bacterial Blight': 'Copper-based fungicides, Streptomycin',
     'Red Rot': 'Fungicides containing Mancozeb or Copper',
@@ -52,12 +56,17 @@ def recommend_pesticide(predicted_class):
         return 'No need for any pesticide, plant is healthy'
     return pesticide_recommendations.get(predicted_class, "No recommendation available")
 
-# Load an H5 Keras model directly.
+# Use st.cache_resource to load H5 models.
 @st.cache_resource(show_spinner=False)
 def load_h5_model(model_path):
-    return tf.keras.models.load_model(model_path, custom_objects={"KerasLayer": hub.KerasLayer}, compile=False)
+    # This will load your H5 Keras model (which contains hub.KerasLayer)
+    return tf.keras.models.load_model(
+        model_path,
+        custom_objects={"KerasLayer": hub.KerasLayer},
+        compile=False
+    )
 
-# Define models dictionary.
+# Define your models dictionary (update paths as needed)
 models = {
     'sugarcane': load_h5_model("models/sugercane_model.h5"),
     'maize': load_h5_model("models/maize_model.h5"),
@@ -66,7 +75,7 @@ models = {
     'wheat': load_h5_model("models/wheat_model.h5"),
 }
 
-# Class names for each model (adjust these to match your models)
+# Class names for each model (ensure these match the order of your model outputs)
 class_names = {
     'sugarcane': ['Bacterial Blight', 'Healthy', 'Red Rot'],
     'maize': ['Blight', 'Common_Rust', 'Gray_Leaf_Spot,Healthy'],
@@ -75,6 +84,7 @@ class_names = {
     'wheat': ['Healthy', 'septoria', 'strip_rust'],
 }
 
+# Preprocess the uploaded image
 def preprocess_image(image_file):
     try:
         image = Image.open(image_file).convert("RGB")
@@ -85,6 +95,7 @@ def preprocess_image(image_file):
         st.error("Error processing image. Please upload a valid image file.")
         return None
 
+# Classify the image using the appropriate model
 def classify_image(model_name, image_file):
     input_image = preprocess_image(image_file)
     if input_image is None:
@@ -95,6 +106,7 @@ def classify_image(model_name, image_file):
     recommended_pesticide = recommend_pesticide(predicted_class)
     return predicted_class, recommended_pesticide
 
+# (Optional) Retrieve detailed plant information from LM Studio
 def get_plant_info(disease, plant_type="Unknown"):
     prompt = f"""
     Disease Name: {disease}