ks-keshava-rao/applefoliarai

Using machine learning to diagnose foliar diseases in apple plants

Apple Foliar AI

Background:

Apple is a widely consumed fruit across the world which has over 7500 cultivars in temperate and subtropical climates. Not only do they hold paramount importance for people's dietary requirements but also drive profitable businesses in several nations. The production and distribution of apples is often hindered by foliar diseases. Currently, the workers on the orchards do scouting to manually identify these diseased leafs and segregate them. For large scale orchards, this eventually turns out to be cumbersome and resource demading. Hence there is a requirement for an automated system to assist in this problem. We propose the use of neural network based algorithms to tackle this issue of diagnosis.

Data:

We would be making use of the Plant Pathology dataset[1] used in the Eight Workshop on Fine-Grained Visual Categorization held in conjuction with CVPR 2021.

The dataset is structured as follows:

|
|-- train images
|-- train.csv : CSV file comprising metadata for training images
      |
      |-- image ID: Basically name of the images
      |
      |-- labels: disease that the apple leaf is diagnosed with
|-- test images: 5003 images
|-- sample_submission.csv: Similar to train.csv. To be populated after training the model on the train images and getting inference on the test images

Our training data is classified into the following categories:

Foliar Disease Condition	Number of Images
Scab	4826
Healthy	4624
frog_eye_leaf_spot	3181
Rust	1860
Complex	1602
Powdery_mildew	1184
scab frog_eye_leaf_spot	686
scab frog_eye_leaf_spot complex	200
frog_eye_leaf_spot complex	165
rust frog_eye_leaf_spot	120
rust complex	97
powdery_mildew complex	87

Usage:

• Clone this repository

git clone https://github.com/smaranjitghose/AppleFoliar.ai.git

• Move inside the AppleFoliar.AI directory

cd AppleFoliar.ai

Setting up code quality and formatting for the project [This is required only when one plans to contribute to the project]

• Install pre-commit package

pip install pre-commit

• Setup pre-commit hooks as per the desired configurations of the project

pre-commit install

• Check if pre-commit is working

pre-commit run

For running inference on a sample image:

python src/inference.py --input path_to_image

Planned work:

• [] Experimentation and Logging of ML models
• [] StreamLit Application
• [] React + SCSS + Bootstrap5 Front
• [] TFJS Inference
• [] Containerization
• [] Documentation

The geeks🤓 behind this initiative:

Our hard-working Project Maintainers👨‍🏫:

Smaranjit Ghose	Anush Bhatia

Our valuable Contributors👩‍💻👨‍💻 :

Documents related to the project:

• LICENSE
• CODE OF CONDUCT
• CONTRIBUTION GUIDELINES
• CHANGELOG

References:

1. Thapa, Ranjita; Zhang, Kai; Snavely, Noah; Belongie, Serge; Khan, Awais. The Plant Pathology Challenge 2020 data set to classify foliar disease of apples. Applications in Plant Sciences, 8 (9), 2020.