Scribd
Upload
18 views12 pages

Plant Disease Detection Using AutoML and Deep Learning 1

The document discusses the development of an automated plant disease detection system using AutoML and deep learning techniques, specifically CNNs, trained on the PlantVillage dataset. The system aims to improve accuracy and efficiency in identifying plant diseases compared to traditional manual inspection methods. Future work includes building a web application for farmers, enhancing model performance, and expanding the detection capabilities to more crops.

Uploaded by

shalbin Lucky 13
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
18 views12 pages

Plant Disease Detection Using AutoML and Deep Learning 1

The document discusses the development of an automated plant disease detection system using AutoML and deep learning techniques, specifically CNNs, trained on the PlantVillage dataset. The system aims to improve accuracy and efficiency in identifying plant diseases compared to traditional manual inspection methods. Future work includes building a web application for farmers, enhancing model performance, and expanding the detection capabilities to more crops.

Uploaded by

shalbin Lucky 13
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 12

Plant Disease

Detection Using
AutoML and Deep
Learning
• Sooraj Rajkumar Poly Raj (200574795)
• Shalbin Kurisuveettil Shaji (200567040)
• Ariga Peruvarathara Vinod (200573386)
• Rithin Ravi (200567791)
• Anjana Kaapaadan(200573241)
🌱 The agricultural industry faces challenges in
plant disease detection affecting crop yields.

🔍 Manual inspection is slow, labor-intensive, and


Problem prone to errors.

Stateme 💡 Deep learning (CNNs) can automate detection,


nt but manual tuning is time-consuming.

🚀 Goal: Build an optimized, automated disease


detection system using AutoML.
✅ Deep learning-based image classification for
plant disease detection.

🔬 Using CNNs trained on the PlantVillage


Solution dataset.

Overview
Model selection
🤖 AutoML (MLJAR) Hyperparameter tuning
automates: Feature optimization

🌍 Final Scope: Web application for real-time


disease detection.
Dataset & Preprocessing
Dataset: PlantVillage (Kaggle)

Labeled images of healthy and diseased plant leaves.

🔄 Preprocessing steps:
Data augmentation (flipping, rotation,
Resizing images to uniform size Normalization for better model training
zoom)
Data Augmentation
Why AutoML (MLJAR)?

🚀 Automates model training and selection.

Model
Selection Optimizes hyperparameters for best performance.

& 📈 Evaluates multiple models and selects the best-performing one.

AutoML
🔹 Process:
1 Input dataset into AutoML
1️⃣
2️⃣AutoML trains multiple models (CNN, Random Forest, XGBoost, etc.)
3️⃣Best model selected automatically based on log loss & accuracy
🏆 Best Performing Model: Ensemble Model
📊 Key Metrics:
🔹 Log Loss: 0.015641

Results & 📈 Accuracy: Highest among tested models


✅ More reliable and generalizable than individual CNNs.
Best 🔬 Performance Evaluation:
Model
Confusion Matrix to analyze classification errors.

Training Curves showing accuracy improvement.


Deployment Plan
🌍 Web Application for Real-World Use

📸 Users: Farmers and agricultural experts


🔗 Workflow:
1 Upload an image of a plant leaf.
1️⃣
2️⃣AI model processes the image.
3️⃣ Output: Disease name & possible
treatment suggestions.
⚡ Benefits:
• Instant results
• Accessible via mobile & desktop
• Reduces reliance on manual inspection
🔹 Key Takeaways:

Successfully implemented AutoML (MLJAR) to automate plant disease detection.

Improved model accuracy using ensemble learning with a log loss of 0.015641.

Addressed challenges in manual tuning by automating model selection and


optimization.

Conclusion 🔹 Impact:

Enhances efficiency in plant disease identification compared to traditional methods.

Provides a strong foundation for real-world agricultural applications.

Sets the stage for deploying a user-friendly web application in Phase 3.


Future Work
1
1️⃣ Phase 3: Web Application Development
• Build a web-based platform for farmers to upload images and receive
disease diagnoses.
• Implement a user-friendly UI for easy access.
2️⃣ Model Improvements
• Train on more diverse datasets to improve real-world generalization.
• Fine-tune AutoML settings for even better accuracy.
3️⃣ Integration of Additional Features
• Provide treatment recommendations based on detected diseases.
• Explore real-time disease monitoring using IoT and edge computing.
4️⃣ Expand to More Crops
• Extend detection beyond the PlantVillage dataset to cover a wider range
of crops.
Thank You!!

You might also like