Performance and Applicability of Transfer Learners for Cocoa Swollen Shoot Detection

Performance and Applicability of Transfer Learners for Cocoa Swollen Shoot Detection

Copyright: © 2021 |Pages: 10
DOI: 10.4018/IJTD.2021040105
OnDemand:
(Individual Articles)
Available
$37.50
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

An accurate and reliable cocoa swollen shoot disease diagnosis is the desire of traditional farmers with low-resolution smart devices. In this study, an efficient cocoa swollen shoot disease identification method base on transfer learners using pre-trained VGG16 and ResNet was proposed. These pre-trained models were trained using 456 samples and validated with 114 samples. The dataset constitutes low-resolution images, VGG16 and ResNet, and achieved an accuracy of 98.25 and 94.73%, respectively. With the objective of proposing a more reliable and accurate model, VGG16 is noted to scale better in terms of performance for implementation.
Article Preview
Top

Introduction

The mother of all civilization is agriculture (Jasim & AL-Tuwaijari, 2020). Agriculture has been practiced since ancient times, and many innovations and developments are taking place (Pooja, Das, & Kanchana, 2017). The most critical driving force for a country's economy is agriculture, which is the livelihood of nearly two thirds of the population of a developing country (Hossain, Hossain, & Rahaman, 2019). Cocoa is a cash crop used for manufacturing various products such as chocolate drinks, bars, cocoa butter, cocoa powder, and among others. Despite all its numerous benefits, cocoa farmers face many challenges in its production. Cocoa is at a high risk of being attacked by pests and diseases in the hot and humid environment. Almost half of the production could be lost due to pest infestation, diseases, and weed. Diseases that can affect yields – most common in West Africa are the black pod and the swollen shoot virus (Ameyaw, Dzahini-Obiatey, & Domfeh, 2014). Plant diseases have contributed to a major decrease in worldwide crop production and quantity (Sun, Zhang, Yang, & Liu, 2020). This has resulted in a drastic reduction in the production and quantity of crops across the world. If these diseases can be precisely detected and treated on time, the economic losses will be significantly minimized, and ecological disasters caused by disease transmission can be prevented (Sun, Zhang, Yang, & Liu, 2020). A significant threat is disease infestation in crops that can negatively impact a nation's economy, including its production and yields, if left unattended to (Francis & Deisy, 2019). Agriculture is one of the most important sources of livelihood that contributes greatly to the economy of the world. However, to a great extent, plant diseases caused by various pathogens can curb yield (Pavel, Rumi, Fairooz, Jahan, & Hossain, 2021). Improving the agricultural product's quality has become critical (Machha, Jadhav, Kasar, & Chandak, 2020). An infestation on crops yield low produce, affecting the consumer's health and the farmers financially as they incur losses (Simon, Kamat, Gutala, & Usmani, 2020). In practice, it can be complicated to manually monitor plant diseases due to an number of factors among which is the processing time. Usually a great effort of experience is required in detecting plant diseases and these expects are mostly few (Khirade & Patil, 2015). When these experts are engaged and become fatigued, many errors are expected to be reported, particularly for some similar leaf diseases (Bi, et al., 2020). In the few decades, the use of smartphones globally and computer vision implementation has allowed smartphone-assisted disease diagnosis (Mohanty, Hughes, & Salathé, 2016). In practice, farmers currently cut down and burn all infected trees to control the infestation; an entire farm can be wiped from the virus. Therefore, there is the need for farmers to detect the early presence of the disease to reduce the number of trees being cut down. The signs of CSSV in the leaves include decolorization of the leaves and veins. Cocoa is a significant cash crop produced in many parts of the world due to its high value as a commodity on the stock market. While many historians agree that Cocoa first originated from Central America, it can now be found in various parts of the world including Africa. Out of the 20 countries ranked as leading producers of Cocoa in the world, eight countries are Africa with Cote D'Ivoire and Ghana being the primary producers. Cocoa production in Africa is known to have contributed nearly 70% percent of the world's Cocoa with an export-import revenue trade-off for small-scale farmers in remote areas. The story in Ghana is not different from other African countries. Numerically, the Cocoa sector has contributed tremendously to Ghana's economy, which offers a means of livelihood for over 700,000 farmers in Ghana's southern tropical belt. Cocoa plays an essential role in the economy of Ghana. The cocoa industry employs approximately 800,00 farmers across six of the ten regions in ghana. The crop generates about $2 billion in foreign exchange annually, and it’s a significant contributor to government revenue and GDP. Therefore, it is expedient to address the issue of CSSV to increase production and reduce the time invested in battling the disease. Research survey has stated that adopting a convolutional neural network in analyzing the images captured by a camera will help minimize the stress these farmers often go through in identifying infected plants and saving farmers time and money. This further has the added advantages of early detection of the infection and reducing the number of infected trees that could have been cut down. In stimulating the complex neuron of the human brain, the deep learning is introduced. These methods has provided several solutions, including plant diseases, to solve real-life problems (Montalbo & Hernandez, 2020). Despite this promising development, little insight is still available into these complex models' activity's internal mechanics or how they achieve such good results. This is hugely unsatisfactory from a scientific perspective (Zeiler & Fergus, 2014). Latest progress in large-scale image and video recognition has been achieved through Convolutional Neural Networks (CNN). Precision agriculture approaches enable stakeholders to make efficient and personalized decisions on crop management regarding the data collected from crop environment monitoring (Chandra, Desai, Guo, & Balasubramanian, 2020). The evolution of technology for image recognition has led to CNN's widespread use in automatic image classification and plant disease recognition (Chen, Liu, & Gao, 2019). With these narratives, there is the need for an engaged research in the field computationally to help with the survival of these cash crop and those who depend on it for livelihold. Visually identifying plant disease as has been the practice of most farmers is inefficient, complicated, time-consuming, costly when it requires agriculture experts in plant diseases, and continuous monitoring might be expensive in large farms. However, with the adoption of computational metods, farmers will be equipped to take surveillance data from the farms and confirm the presence of swollen shoot disease.

Complete Article List

Search this Journal:
Reset
Volume 15: 1 Issue (2024): Forthcoming, Available for Pre-Order
Volume 14: 1 Issue (2023)
Volume 13: 4 Issues (2022): 1 Released, 3 Forthcoming
Volume 12: 4 Issues (2021)
Volume 11: 4 Issues (2020)
Volume 10: 4 Issues (2019)
Volume 9: 4 Issues (2018)
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing