import gradio as gr
import torch
from PIL import Image
from torchvision import transforms
import warnings
import sys
import google.generativeai as genai
import os
import contextlib
from transformers import ViTForImageClassification, pipeline
# Suppress warnings related to the model weights initialization
warnings.filterwarnings("ignore", category=UserWarning, message=".*weights.*")
warnings.filterwarnings("ignore", category=FutureWarning, module="torch")
# Suppress output for copying files and verbose model initialization messages
@contextlib.contextmanager
def suppress_stdout():
with open(os.devnull, 'w') as devnull:
old_stdout = sys.stdout
sys.stdout = devnull
try:
yield
finally:
sys.stdout = old_stdout
# Load the saved model and suppress the warnings
with suppress_stdout():
model = ViTForImageClassification.from_pretrained('google/vit-base-patch16-224-in21k', num_labels=6)
model.load_state_dict(torch.load('vit_sugarcane_disease_detection.pth', map_location=torch.device('cpu')))
model.eval()
# Define the same transformation used during training
transform = transforms.Compose([
transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
])
# Load the class names (disease types)
class_names = ['BacterialBlights', 'Healthy', 'Mosaic', 'RedRot', 'Rust', 'Yellow']
#Gemini Response
def get_response_llm(predicted_label,knowledge_base):
prompt = f"Your an helpful assistant who helps farmers know about the sugarcane leaf diseases , precaution, advise etc....Predicted disease label will is given to you '{predicted_label}' and also {knowledge_base} Provide breif answer of advise for managing this condition.Give the response in a beautiful way like bold or bullet point etc.. wherever required"
genai.configure(api_key=os.getenv("GEMINI_API_KEY"))
model = genai.GenerativeModel("gemini-1.5-flash")
response = model.generate_content([prompt])
return response.text
# Comprehensive knowledge base for sugarcane diseases and practices
knowledge_base = """
'BacterialBlights': Bacterial blights are caused by *Xanthomonas albilineans*.
**Symptoms:**
- Water-soaked lesions on leaves.
- Gradual yellowing and withering of leaves.
- Reduction in photosynthesis and stunted growth.
**Management:**
- Apply copper-based fungicides.
- Improve field drainage to avoid waterlogging.
- Use disease-free planting material.,
'Mosaic': Mosaic disease is caused by the Sugarcane mosaic virus (SCMV) and often transmitted by aphids.
**Symptoms:**
- Mottled appearance on leaves with streaks of yellow and green.
- Reduced photosynthetic efficiency.
- Decreased cane weight and sugar content.
**Management:**
- Use resistant sugarcane varieties.
- Control aphid populations with insecticides.
- Remove and destroy infected plants to prevent spread.,
'RedRot': Red rot is caused by the fungus *Colletotrichum falcatum*.
**Symptoms:**
- Red streaks inside the cane with white patches.
- Rotting of the stalk, emitting a sour smell.
- Drying of leaves and eventual plant death.
**Management:**
- Plant resistant varieties.
- Remove and burn infected plants.
- Treat soil with fungicides and practice crop rotation.,
'Rust': Rust is caused by the fungus *Puccinia melanocephala*.
**Symptoms:**
- Formation of orange to reddish pustules on leaves.
- Premature drying of leaves.
- Reduced plant vigor and yield.
**Management:**
- Apply systemic fungicides (e.g., triazoles).
- Ensure proper field hygiene.
- Avoid water stress and maintain balanced nutrition.,
'Yellow': Yellowing can be caused by nutrient deficiencies or disease onset.
**Symptoms:**
- Yellowing of leaf tips or entire leaves.
- Reduced photosynthesis and growth.
**Management:**
- Conduct soil testing to identify deficiencies.
- Apply balanced fertilizers as per soil nutrient status.
- Maintain proper irrigation schedules.,
'Smut': Smut is caused by the fungus *Sporisorium scitamineum*.
**Symptoms:**
- Formation of whip-like structures at the growing points.
- Stunted growth and tiller proliferation.
- Reduced sugar content.
**Management:**
- Plant smut-resistant varieties.
- Remove smut-infected plants.
- Treat seed sets with fungicides before planting.,
'Healthy': The sugarcane crop is healthy.
Continue regular monitoring and follow good agronomic practices:
- Ensure balanced fertilization.
- Maintain proper irrigation schedules.
- Monitor for pests and diseases regularly.,
'GeneralPractices': **General Practices for Disease Prevention**
- **Field Sanitation:** Remove and destroy crop residues and infected plants to reduce inoculum levels.
- **Resistant Varieties:** Cultivate sugarcane varieties that are resistant to specific diseases.
- **Seed Treatment:** Use disease-free, certified seed material. Treat seed sets with fungicides before planting.
- **Crop Rotation:** Rotate sugarcane with non-host crops to break the disease cycle.
- **Optimal Agronomic Practices:** Ensure proper irrigation and drainage. Maintain balanced fertilization and avoid over-application of nitrogen.
- **Timely Monitoring and Control:** Inspect fields regularly for symptoms. Apply recommended fungicides or bactericides as soon as symptoms appear.
- **Integrated Pest and Disease Management (IPDM):** Combine biological, chemical, and cultural methods to manage diseases sustainably.,
'ImpactOfDiseases': **Impact of Sugarcane Diseases**
- **Yield Reduction:** Diseases like red rot and smut can reduce cane yield by 30–60%.
- **Quality Degradation:** Affected plants produce less sugar and lower-quality juice.
- **Economic Losses:** Increased cost of management and reduced marketable output affect profitability.,
'SugarcaneOverview': Sugarcane is a critical crop globally, providing raw materials for sugar, ethanol, and other byproducts.
However, it is susceptible to various diseases caused by fungi, bacteria, viruses, and environmental factors.
Effective management practices are essential to ensure high yield and quality.
"""
# Update the predict_disease function
def predict_disease(image):
# Apply transformations to the image
img_tensor = transform(image).unsqueeze(0) # Add batch dimension
# Make prediction
with torch.no_grad():
outputs = model(img_tensor)
_, predicted_class = torch.max(outputs.logits, 1)
# Get the predicted label
predicted_label = class_names[predicted_class.item()]
# # Retrieve response from knowledge base
# if predicted_label in knowledge_base:
# detailed_response = knowledge_base[predicted_label]
# else:
# # Fallback to AI-generated response
predicted_label = f'The predicted label is {predicted_label}'
detailed_response = get_response_llm(predicted_label,knowledge_base)
# Create a styled HTML output
output_message = f"""
Detected Disease: {predicted_label}
"""
if predicted_label != "Healthy":
output_message += f"""
{detailed_response}
"""
else:
output_message += f"""
{detailed_response}
"""
return output_message
# Create Gradio interface
inputs = gr.Image(type="pil")
outputs = gr.HTML() # Use HTML output for styled text
EXAMPLES = ["img1.jpeg", "redrot2.jpg", "cropped_yellow.jpeg","cropped_rust.jpeg", "cropped_BacterialBlight.png","cropped_mosaic.jpeg","healthy2.jpeg"]
demo_app = gr.Interface(
fn=predict_disease,
inputs=inputs,
outputs=outputs,
title="Sugarcane Disease Detection",
examples=EXAMPLES,
live=True,
theme="huggingface"
)
demo_app.launch(debug=True)