Download Resume
Tomato Leaf Disease Detection

Tomato Leaf Disease Detection

(2024)

AI-Powered Crop Health Monitoring

AI/MLComputer VisionAgriculture
☕☕☕☕
Coffee Required
Senior
Complexity
⭐⭐⭐⭐
Documentation
👍👍
Would Build Again

Duration: 8 weeks

Synopsis

Custom YOLO-11M object detection model for automated identification of 7 distinct tomato leaf diseases. Implemented Fourier-based regularization for sharper disease localization.

The Making Of

🎯 The Challenge

Manual crop inspection is time-consuming and error-prone. Need to identify 7 distinct leaf states including Bacterial Spot, Early Blight, and Late Blight with precise localization.

💡 The Solution

Built custom YOLO-11M architecture with C2f Fusion Blocks (25% parameter reduction), SPPF for multi-scale pooling, and Polarized Self-Attention. Innovated with Fourier Loss term using 2D FFT to preserve high-frequency edge details.

🚀 The Impact

Fourier-enhanced model improved Box Precision by 2.4% and mAP@50 by 2.4% over baseline with zero inference overhead. Model focuses on disease textures rather than being distracted by shadows or dirt.

Key Metrics

78.16%
Precision
83.28%
mAP@50
~22ms
Inference Speed

Tech Stack (Cast & Crew)

  • YOLO-11MObject Detection
  • PyTorchML Framework
  • FFTFourier Loss
  • PythonImplementation

Quick Info

Year
2024
Duration
8 weeks
Complexity
Senior
Featured
Yes
← Back to All Projects