Plant_Disease_Classification_CNN

Documents

Report Code
PDF Jupyter Notebook
HTML Kaggle Notebook

Extended Abstract

Problem

Misdiagnosis of the many diseases impacting agricultural crops can lead to misuse of chemicals leading to the emergence of resistant pathogen strains, increased input costs, and more outbreaks with significant economic loss and environmental impacts. Current disease diagnosis based on human scouting is time-consuming and expensive, and although computer-vision based models have the promise to increase efficiency, the great variance in symptoms due to age of infected tissues, genetic variations, and light conditions within trees decreases the accuracy of detection.

Objectives

The Plant Pathology Challenge we have attended consists in training a model using images of the training dataset to

Submissions are evaluated on mean column-wise ROC AUC.

Data

Both the training and the testing datasets are composed of 1821 high-quality, real-life symptom images of multiple apple foliar diseases to be classified into four categories: healthy, multiple_diseases, rust, scab.

Methods

  1. Class balancing with SMOTE
  2. Data augmentation with Keras ImageDataGenerator
  3. Learning rate schedule
  4. Optimal dropout
  5. Epoch grid search
  6. Visualization of convolutional filters and activation maps of the layers.

Results

Model ROC AUC
Keras DenseNet121 0.972
InPhyT EKM 0.937

References

  1. Plant Pathology 2020 - FGVC7: Identify the category of foliar diseases in apple trees, Kaggle (2020).
  2. Ranjita Thapa et al. The Plant Pathology 2020 challenge dataset to classify foliar disease of apples, arXiv pre-print (2020).