(IJACSA) International Journal of Advanced Computer Science and Applications Vol. 14, No. 9, 2023 235 | Page www.ijacsa.thesai.org Improved YOLO-X Model for Tomato Disease Severity Detection using Field Dataset Rajasree R, C Beulah Christalin Latha Department of Digital Sciences, Karunya Institute of Technology and Sciences, Coimbatore, India Abstract—In the past decade, the field of automatic plant disease detection has undergone significant complexity. Advancements in convolutional neural network in deep learning have enabled the rapid and precise detection of ailments, facilitated the development of effective treatments and ultimately led to higher crop yields. One of the most challenging scenarios in plant disease occurs when multiple diseases manifest on a single leaf, exacerbating the difficulty of diagnosis due to overlapping symptoms. This study addresses these challenges by employing an enhanced YOLO-X model for detection tomato leaf diseases. The technique presented here enhances the Spatial Pyramid Pooling layer in order to extract valuable features from training data of various sizes more efficiently. We were able to increase the model's ability to identify a broader spectrum of disease symptoms by concatenating variables from multiple layers and varying sizes. In addition, we incorporate a large number of connections to increase the generalizability of the design. The application of an IoU-based (Intersection over Union) regression loss function increases the convergence of the network and the precision of the detection. For experimentation, we created a customized dataset consisting of 1220 tomato plant leaf images from various farms in Southern part of India, encompassing overlapping diseases and varying degrees of severity. The dataset includes images of healthy leaves as well as different severity levels of tomato leaf curl and tomato leaf mold stress on a single leaf. Our suggested improved SPP-based YOLO-X model beats the original YOLO-X model, according to experimental findings, which show an improvement in test dataset accuracy and a 73.42% mean Average Precision on field-collected dataset. Keywords—Convolutional neural network; deep learning; object classification; plant disease detection; spatial pyramid pooling; YOLOX I. INTRODUCTION In India, tomatoes are a significant economical crop that is produced on 15% of the nation's total cultivated land. A significant portion of the global textile economy is contributed by the nation's tomato production and export, in addition to its local consumption. The crop is afflicted with several diseases during the course of its existence. A leaf might sometimes have many diseases, some of which have similar symptoms. Even an experienced pathologist may make mistakes when evaluating disease severity signs and the presence of numerous stressors. Precision farming practices have undergone a revolution with the development of artificial intelligence and computer vision technology. In plant disease detection systems, a number of machine learning and deep learning models have shown outstanding performance [1] on field-collected or publicly accessible plant disease datasets, some researchers have combined deep learning-based feature extraction and classification tasks with transfer learning [8]. In order to propose the use of pesticides or other preventative measures and achieve near-ideal performance in recognizing diseases signs automatically, several research have been conducted. Well-known deep learning architectures such as region-based convolutional networks [2], single shot detectors [3], and region proposal networks [4] have been employed in the area of plant leaf disease detection, with major alterations happening during the preceding few years [37, 38, 41]. Almost all previous studies either used the well-known PlantVillage public dataset [3] or their own datasets collected in the field [5], [6]. But only a small number of studies have looked at the stages of disease growth and the chance that many living and nonliving things can attack a plant leaf at the same time. In these situations, it is hard for both human and automatic monitoring systems to figure out the type of infection and the exact area of sickness signs. In this study, we describe a YOLO-X-s based detecting system that uses a modified Spatial Pyramid Block to combine fine spatial data with local features to find sickness phases and split diseases with symptoms that overlap. We made the spatial pyramid pooling block better by putting together feature maps at low-level scales. This helped us solve the problem more accurately. The original size feature vector was added to improve the quality of the features. The recognition performance got even better when the Alpha IoU regression loss function was used [36]. A. Contributions to the Research  A better YOLO-Xs model with a modified Pyramid pooling module (SPP) layer is given so that many diseases on a single leaf plant can be found. It collects location information at local, multi-scale levels to get the information it needs more quickly.  To improve generalization and convergence, we used Alpha IoU (Intersection over Union) loss as the bounding box regression for multiple disease localization when multiple diseases showed up on the same plant leaf.  With the help of enhancement, a group of unique shots from a tomato field are shown. The photos show how diseases spread and how many different diseases can be found on a single leaf. The paper has been structured in the following manner: The Section II provides a summary of the existing literature. The Section III describes the proposed methodology. The Section IV presents research outcomes, comparisons to existing