A low-cost, fully autonomous agricultural rover designed to detect tomato plant diseases in real-time using deep learning.
This project integrates EfficientNetB0, Raspberry Pi 5, and a Pi Camera Module 2, along with a Telegram bot for instant alert notifications.
- 🚜 Autonomous rover navigating through crop fields
- 📷 Real-time image capture using Raspberry Pi Camera Module
- 🤖 EfficientNetB0 deep learning model deployed directly on Raspberry Pi 5
- 📡 Offline inference, no cloud / internet dependency
- 🔔 Telegram alerts for detected diseases
- 🌾 Trained on PlantVillage dataset
- 🎯 Achieved 99.73% accuracy
PlantVillage Tomato Leaf Disease Dataset
🔗 https://www.kaggle.com/datasets/mohitsingh1804/plantvillage
Dataset details:
- 16,012 original tomato leaf images
- 10 classes (9 diseases + 1 healthy)
- Added 902 background images
- Total images used → 16,914
| Model | Test Accuracy |
|---|---|
| EfficientNetB0 | ⭐ 99.73% |
| ResNet18 | 98.13% |
| MobileNetV2 | 96.35% |
EfficientNetB0 was selected due to exceptional accuracy & fast inference on Raspberry Pi 5.
Rover → Camera → Raspberry Pi 5 → EfficientNetB0 Model → Telegram Alert → Output
- Raspberry Pi 5
- Raspberry Pi Camera Module 2
- L298N Motor Driver
- 4-Wheel Rover Base
- Li-ion Battery / Power Bank
- Jumper Wires
Python 3.10+
TensorFlow / Keras
OpenCV
NumPy
Pillow
Telegram Bot API
Install dependencies: pip install -r requirements.txt
Outputs include:
- Disease class
- Confidence score
- Timestamp
- Leaf image
Set your bot token inside: config/telegram_config.py
The rover was tested under:
- Sunlight
- Shadows
- Varying leaf positions
- Natural environmental noise
The model produced stable real-time predictions with high accuracy.
✔ Raspberry Pi 5 + EfficientNetB0 for real-time inference
✔ Added background class to reduce false positives
✔ Fully autonomous rover navigation system
✔ Telegram alert pipeline for farmers
- GPS-based disease location mapping
- YOLO-based leaf localization
- Integration with soil sensors
- Multi-crop disease detection
- Drone + Rover hybrid system
Shree Santh B Aditya S Padmacharan R Ashwin Naresh M Amrita School of Artificial Intelligence
MIT License
If using or referencing this project:
“A Convolutional Neural Network Approach for Real-Time Tomato Disease Classification on a Mobile Robotic Platform, 2024.”