Leaf Disease Classification using Deep Learning

Local Chapter Addis Ababa, Ethiopia Chapter

Coordinated by,

Status: Completed

Project Duration: 01 Apr 2023 - 31 May 2023

Open Source resources available from this project

Project background.

Faba Bean (Vicia Faba L.), which is known as bak’ēla locally in Amharic, is one of the earliest domesticated grain legumes in the world. For countries like Ethiopia, it is highly consumed domestically and serves as a major source of protein for humans and animals. In addition, it is also one of the popular sources of income for farmers and the country by availing it in the local and export market. The other contribution of Faba Bean is its soil fertility restoration for farmlands during crop rotation and also balancing the atmospheric nitrogen (Gedyon Tamiru, 2017).

Ethiopia is the second-largest producer of Faba Bean according to the FAO-STAT 2016 report (Beyene Bitew & Tigabie A., 2016; Bogale Nigir et al., 2016). However, yield in the country is generally very low due to production constraints mainly disease (Anteneh Ademe et al., 2018). Diseases are a major threat to the production and quality of agricultural plants such as Faba Bean.
According to the EFDRE Ministry of Agriculture pulse crop manual (Ethiopian Institute of Agricultural Research, 2016) among the various diseases that affect the Faba Bean crop, Chocolate Spot (Botrytis Fabae), Rust (Uromyces Fabae), Gall disease (Olpidium Viciae) and Aschochyta Blight (Aschochyta Fabae) and Fusarium Solani are the common ones. Once those diseases appear on the Faba Bean plant, visible symptoms begin to reveal on the leaf organ of a Faba Bean plant.

Chocolate spot (Botrytis fabae) is a common disease that occurs almost everywhere Faba Bean is grown (Kumar, 2018). Its symptoms vary from small spots on the leaves (Figure 1-A) to the complete blackening of the entire plant. Though leaves are the main parts of the plant its symptoms are shown, it also spreads to stems, flowers, and pods under favorable conditions. The infection starts as small reddish-brown circular spots on leaves, stems, and flowers. The spots on leaves and stems enlarge and develop a grey, dead center with a red-brown margin.

Rust (Uromyces Fabae) is also another Faba Bean disease which is widely distributed in the Faba Bean production areas, but severe in humid tropical and subtropical areas. It is first seen on the leaves of the plant as rusty red pustules surrounded by a light yellow halo that densely cover the surface of the leaf, and later distributed to the stems (Kumar, 2018). The creamy-yellow spots or pustules on leaves (Figure 1-B.) are clearly visualized on the Faba Bean leaf which helps to identify the disease.

The problem.

These days, many researchers are conducting research and applying deep learning technologies to different plant disease detection problems. Faba Bean diseases are the major challenges of the crop producers which causes a huge loss in both quality and quantity. Analyzing the symptoms seen on the leaves of the plant and recognizing the disease play a key role in the successful cultivation of crops. A proper diagnosis mechanism is an extremely important task in preventing the disease in the early stages. Proper identification of diseases and taking immediate control measures are the most important aspects of plant pathology.

Through our best knowledge, no automated Faba Bean disease detection system is designed either using traditional machine learning or deep learning techniques. Even though many crop disease detection systems are developed by applying those systems for Faba Bean disease detection is not feasible. This is because the disease features such as color and texture are different from crop to crop. Therefore, we decided to develop a Faba Bean crop leaf disease detection system by applying a deep learning technique called CNN.

Project goals.

- Data Collection and Exploratory Data Analysis - Preprocessing  - Feature Extraction - Model Development and Training - Evaluate Model - Prototype development

Project plan.

  • Week 1

    Researching previous works and how to proceed with the Data Collection

  • Week 2

    Data Collection

  • Week 3

    Exploratory Data Analysis

  • Week 4

    Preprocessing and Augmentation

  • Week 5

    Model Development

  • Week 6

    Model Training

  • Week 7

    Model Analysis and Interpretation

  • Week 8

    Prototype Development

Learning outcomes.

Image processing, preprocessing, model building and testing.

Share project on: