DeepSeqCoco: A Robust Mobile Friendly Deep Learning Model for Detection of Diseases in Cocos nucifera

Miit Daga, Dhriti Parikh, Swarna Priya Ramu · 2025

Read DeepSeqCoco: A Robust Mobile Friendly Deep Learning Model for Detection of Diseases in Cocos nucifera by Miit Daga, Dhriti Parikh, Swarna Priya Ramu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-10030.txt.

Introduction

Coconut growers in developing countries often spot tree diseases late, since checks are manual and slow. Miit Daga, Dhriti Parikh, and Swarna Priya Ramu built a phone-friendly deep learning model in 2025 that names coconut diseases from tree images.

Their model, DeepSeqCoco, was trained and tested under several optimizer settings, including a hybrid that mixes two standard training rules. The hybrid gave the lowest validation loss, and the model reached accuracy as high as 99.5 percent, up to five points above earlier models. It also trained up to 18 percent faster and answered up to 85 percent faster per image. The authors pitch it as a cheap, scalable tool for early disease warnings on farms.

Catalog record

CREATORMiit Daga, Dhriti Parikh, Swarna Priya Ramu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE71 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-10030

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

DeepSeqCoco: A Robust Mobile Friendly Deep Learning Model for Detection of Diseases in Cocos nucifera Miit Daga, Dhriti Parikh, Swarna Priya Ramu Abstract Coconut tree diseases are a serious risk to agricultural yield, particularly in developing countries where conventional farming practices restrict early diagnosis and intervention. Current disease identification methods are manual, labor-intensive, and non-scalable. In response to these limitations, we come up with DeepSeqCoco, a deep learning based model for accurate and automatic disease identification from coconut tree images. The model was tested under various optimizer settings, such as SGD, Adam, and hybrid configurations, to identify the optimal balance between accuracy, minimization of loss, and computational cost. Results from experiments indicate that DeepSeqCoco can achieve as much as 99.5% accuracy (achieving up to 5% higher accuracy than existing models) with the hybrid SGD-Adam showing the lowest validation loss of 2.81%. It also shows a drop of up to 18% in training time and up to 85% in prediction time for input images. The results point out the promise of the model to improve precision agriculture through an AI-based, scalable, and efficient disease monitoring system. I Introduction Coconut tree is one such plant whose all parts, from the fruit to the trunk are useful in some or the other way. It is also known as the "tree of life" as it is used for food, medicine, fuel and so on. However, coconut cultivation faces significant challenges in various countries around the world, most of which are either under developed or developing nations. In Côte d’Ivoire (a country in West Africa), an outbreak resulted in destruction of 350 hectares of coconut plantation and caused a loss of 12,000 tons of copra each year [ 1 ] . In the South-east Asian country of Philippines, the second largest producer of coconut in the world, a 2012-2013 Philippine Coconut Authority (PCA) survey reported more than half a million Cadang-Cadang-infected palms in the country. According to their estimate, at a spread rate of 0.1-1% annually, by 2032, around 3 million coconut trees would be at a risk of infection. If not intervened in timely, it could result in losses of approximately 300 million dollars by 2032 [ 2 ] . India, the third largest coconut producer faces yield losses ranging from 10% to 80% due to diseases like Root Wilt Disease (RWD), Leaf Blight, Bud Rot and Basal Stem Rot [ 3 ] . In Sri Lanka, infestation due to white flies have led to a 40% reduction in king coconut output with more than 4 million affected coconut palms [ 4 ] . To address these issues, the past decade has seen extensive advances in the field of smart agriculture, especially in precision agriculture [ 5 ] . The reason behind this is that even though there is a proportional decrease in the amount of agricultural land available, there is an increasing demand for food across the world due to an increased population. To handle this growing demand, the entire agricultural process should be mechanized. When using such a process mechanization as a remedy, the machines are made to perform most of the functionalities of the agricultural process there by increasing the effectiveness and yield [ 6 ] . In the case of developing countries, the utilization of machines for agricultural processes is limited and inefficient, they instead focus on automating the entire functionalities of farming to improve the rural capabilities. In order to address these challenges, modern day Artificial Intelligence (AI) powered solutions can be handy in providing real-time disease detection and help farmers especially in developing and under developed nations where age-old farming methods are still prevalent. This paper proposes a Deep Learning (DL) based mobile friendly framework DeepSeqCoco model, an intelligent solution to effectively identify coconut tree diseases through images. Utilizing powerful DL algorithms, it facilitates early and accurate diagnosis, enabling timely action and limiting crop loss, particularly in areas where conventional monitoring procedures are ineffective. II Related Works Recent years have seen multiple researchers focus on societal problems and try to provide solutions, including agriculture as an important domain. The advancements in AI, Machine Learning (ML) and DL have paved the way for new methodologies in predicting the yield, pest detection, disease detection, and providing precautions to increase the production and hence the well-being of the farmers. This section discusses in detail the recent work carried out by various researchers around the world in the agriculture domain, especially, disease prediction in coconut trees. Anitha et al. proposed a methodology for detecting the leaf disease in coconut trees using DenseNet 121 architecture. They performed pre-processing on the data set for handling the difference in color, orientation and shape of the leaves. ReLU activation function was utilized for enhancing the feature propagation. At the end of every epoch, a mechanism was used for automatically adjusting the learning rate. They validated their data set using the F1 score, accuracy, recall, and precision as evaluation parameters and achieved an accuracy of 99% [ 7 ] . Bharathi et al. came up with a methodology for early detection of diseases in coconut tree leaves using K-means clustering, statistical, Cubic Support Vector Machine (SVM) and gray-level co-occurrence matrix (GLCM). While k-means clustering took care of segmentation, feature extraction was done through GLCM and Cubic SVM helped in classification. Data pre-processing was done using two main techniques: Contrast Enhancement and Histogram Equalization. While many features were extracted, such as mean, contrast, homogeneity, energy, entropy, variance, smoothness, standard deviation, and skewness; only mean, contrast, and energy were used with cubic SVM, achieving the highest accuracy of 97.3% [ 8 ] . Nesarajan et al. devised a model using image processing and DL techniques for prediction of nutrient deficiencies and pest diseases in coconut trees. They used SVM for classification of leaves and Convolutional Neural Networks (CNN) to predict the actual nutrient deficiency. While they tested multiple CNN architectures like EfficientNetB0, ResNet50 and VGG16, they finally decided upon EfficientNetB0 as it demonstrated the best accuracy of 93.72% when compared to other architectures for prediction of nutrient deficiency. They also tested multiple algorithms including SVM, logistic regression and k-Nearest Neighbour (KNN) for pest detection but SVM outperformed all of them with an accuracy of 93.54%. For this, they extracted features like color, shape and skewness from the images. K-means clustering technique was used for enhancing leaf image quality [ 9 ] . Vidhanaarachchi et al. developed a model for identifying coconut disease and pest infestations using DL architectures. They resized all the images in the dataset to a dimension of 300x300 and employed pixel normalization (dividing by 255) before using them with DL models like DenseNet-121, DenseNet-169, InceptionResNetV2, etc. They used Adam optimizer along with a learning rate ranging from 0.003 to 0.007 during the training phase. After testing multiple models, they finally settled with DenseNet-121 for Weligama Coconut Leaf Wilt Disease (WCWLD) detection as it gave 99% training accuracy and 90% testing accuracy. On the other hand, to determine the severity of symptoms associated with WCWLD, InceptionResNetV2 was used which achieved 99.69% training accuracy and testing accuracy of 97%. DenseNet-121 was also used for detecting Leaf Scorch Decline (LSD) which gave a training accuracy of 98.78% and testing

Read the complete text