Plants-Disease-Detection_using_TF_with_Keras

Plants Disease Detection using CNN Model in Machine Learning.

List of contents

• Introduction
• Approach
• Dataset Description
• Results

Introduction

(Back to top)

Plant disease can directly lead to stunted growth causing bad effects on yields. An economic loss of up to $20 billion per year is estimated all over the world. Diverse conditions are the most difficult challenge for researchers due to the geographic differences that may hinder the accurate identification. In addition, traditional methods mainly rely on specialists, experience, and manuals, but the majority of them are expensive, time-consuming, and labor-intensive with difficulty detecting precisely. Therefore, a rapid and accurate approach to identify plant diseases seems so urgent for the benefit of business and ecology to agriculture.

In agriculture products, diseases are the main cause for the lessening in both quality and production of the agriculture products. Farmers puts their great effort in picking best seeds of plant and also provide proper environment for the growth of the plant, although there are lot of diseases that affects plant result in plant disease.

Recognition of the deleterious regions of plants can be considered as the solution for saving the reduction of crops and productivity. The past traditional approach for disease detection and classification requires enormous amount of time, extreme amount of work and continues farm monitoring.

Aproach

(Back to top)

1. Choice of deep learning architecture:

• GoogLeNet(Inception_V3)

2. Choice of training mechanism:

• Transfer Learning

3. Choice of dataset type:

• Color

4. Choice of training-testing set distribution:

• Train: 80%, Test: 20%
  

The inception module uses parallel 1 × 1, 3 × 3, and 5 × 5 convolutions along with a max-pooling layer in parallel, hence enabling it to capture a variety of features in parallel.

Dataset Description:

(Back to top)

Plant Village dataset is a public dataset of 54,305 images of diseased and healthy plant leaves collected under controlled conditions ( PlantVillage Dataset). The images cover 14 species of crops, including: apple, blueberry, cherry, grape, orange, peach, pepper, potato, raspberry, soy, squash, strawberry and tomato. It contains images of 17 basic diseases, 4 bacterial diseases, 2 diseases caused by mold (oomycete), 2 viral diseases and 1 disease caused by a mite. 12 crop species also have healthy leaf images that are not visibly affected by disease.

Name	No of Classes	Class Names
Apple	04	'Apple___Apple_scab','Apple___Black_rot','Apple___Cedar_apple_rust' 'Apple___healthy'
Blueberry	01	'Blueberry___healthy'
Cherry	02	'Cherry_(including_sour)Powdery_mildew', 'Cherry(including_sour)_healthy'
Corn	04	'Corn___Cercospora_leaf_spot', 'Corn___Common_rust','Corn___Northern_Leaf_Blight','Corn___healthy'
Grape	04	'Grape___Black_rot','Grape___Esca_(Black_Measles)','Leaf_blight_(Isariopsis_Leaf_Spot)','Grape___healthy'
Orange	01	'Orange___Haunglongbing_(Citrus_greening)'
Peach	02	'Peach___Bacterial_spot','Peach___healthy'
Pepper	02	'Pepper,_bell___Bacterial_spot','Pepper,_bell___healthy'
Potato	03	'Potato___Early_blight','Potato___Late_blight','Potato___healthy'
Raspberry	01	'Raspberry___healthy'
Soyabean	01	'Soybean___healthy'
Squash	01	'Squash___Powdery_mildew'
Strawberry	02	'Strawberry___Leaf_scorch','Strawberry___healthy'
Tomato	10	Tomato: 'Bacterial_spot','Early_blight', 'Late_blight', 'Leaf_Mold', 'Septoria_leaf_spot', 'Spider_mites','Target_Spot', 'Yellow_Leaf_Curl_Virus', 'Mosaic_virus', 'Healthy'

Results

(Back to top)

Acieved an training accuracy of 90.19 and validation accuracy of 91.97