Abstract:As global food demands escalate, ensuring optimal crop health has become paramount. Traditional disease detection methods often fall short in terms of speed and accuracy, emphasizing the need for advanced, technology-driven solutions. Thus, this paper explores the impact of fine-tuning pre-trained CNN architectures synchronously with the structure of CNN model, focusing on disease detection in tomato leaves. We suggested utilizing these pre-trained CNN architectures as a feature extraction phase and tuning them alongside the classification phase of the proposed model. We posit that the harmonious tuning of both the feature extraction phase and the classification phase holds the potential to enhance the performance of any deep-learning model. In pursuit of this objective, we extended the concept of hyperparameters to encompass the feature extraction phase, alongside a comprehensive spectrum of hyperparameters that influence deep-learning models. Subsequently, we harnessed the random search algorithm to optimize these hyperparameters and determine the optimal model architecture for enhanced tomato disease detection accuracy. The model refined through the random search algorithm was designated as Xception-CNN. Initially, we trained and evaluated the proposed Xception-CNN model using the tomato leaves dataset. Subsequently, we conducted an experiment by removing the feature extraction phase from the Xception-CNN model and transforming it into an end-to-end scratch-CNN model. This step aimed to both validate the efficacy of our approach and unveil the impact of fine-tuned pre-trained model feature maps on CNN model performance. The outcomes indicated the superior performance of the proposed Xception-CNN model compared to the Scratch-CNN model across all evaluation metrics. Notably, the classification accuracy of the Scratch-CNN model peaked at 76.70%, whereas the Xception-CNN model achieved an impressive accuracy of 99.40%. These findings underscore the significance of meticulous deep-learning model refinement, coupled with the utilization of pre-trained models, within such an optimization and refinement process.

Evaluating the impact of tuned pre-trained architectures' feature maps on deep learning model performance for tomato disease detection

An efficient deep learning model for tomato disease detection

Impact of Pretrained Deep Neural Networks for Tomato Leaf Disease Prediction

Design of Efficient Methods for the Detection of Tomato Leaf Disease Utilizing Proposed Ensemble CNN Model

On the Image-Based Detection of Tomato and Corn leaves Diseases : An in-depth comparative experiments

Comparison of Convolutional Neural Network Architectures for Classification of Tomato Plant Diseases

Early Detection and Classification of Tomato Leaf Disease Using High-Performance Deep Neural Network

Diagnosis of tomato pests and diseases based on lightweight CNN model

FINE-GRAINED TOMATO DISEASE RECOGNITION BASED ON DEEP CONVOLUTIONAL NETWORK

Revolutionizing tomato disease detection in complex environments

Transform and Deep Learning Algorithms for the Early Detection and Recognition of Tomato Leaf Disease

Enhancing Plant Disease Detection: A Novel CNN-Based Approach with Tensor Subspace Learning and HOWSVD-MD

Detection and Classification of Tomato Crop Disease Using Convolutional Neural Network

An image based prediction system for early blight disease in tomato plants using deep learning algorithm

Tomato plant leaf diseases detection and classification using an advanced self-developed CNN model

A Lightweight Attention-Based Convolutional Neural Networks for Tomato Leaf Disease Classification

High-Performance Deep Neural Network-Based Tomato Plant Diseases and Pests Diagnosis System With Refinement Filter Bank

AlexNet Convolutional Neural Network for Disease Detection and Classification of Tomato Leaf

Precision Agriculture Through Deep Learning: Tomato Plant Multiple Diseases Recognition With CNN and Improved YOLOv7

Automated Disease Diagnosis in Pumpkin Plants Using Advanced CNN Models