Authors:
M. Arunadevi Thirumalraj, B. Rajalakshmi, B. Santosh Kumar, S. Gopikha, Amarilys González García

Addresses:
Department of Computer Science and Engineering, Karunya Institute of Technology and Science, Coimbatore, Tamil Nadu, India. Department of Computer Science and Business Management, Saranathan College of Engineering, Tiruchirappalli, Tamil Nadu, India. Department of Computer Science and Engineering, New Horizon College of Engineering, Bengaluru, Karnataka, India. Department of Information Technology, St. Joseph's College of Engineering, Chennai, Tamil Nadu, India. Department of Research and Development, Placental Histotherapy Center, Havana, Cuba.

Abstract:

Automating fruit detection is a continuous challenge due to its complexity. Because fruit varieties and subtypes may vary by geography, manually classifying fruits can be challenging. The Fruit-360 dataset was categorised using convolutional neural network-based techniques (e.g., VGG16, Inception V3, MobileNet, and ResNet-18) in several recent publications. Unfortunately, the 131 fruit classifications are not comprehensive enough to be of much service. Furthermore, the computational efficiency of these models was poor. With 90,483 sample images and 131 fruit categories, our innovative, comprehensive, and reliable study can recognise and predict them. A modified AlexNet-based strategy, combined with an effective classifier, was employed to bridge the research gap effectively. The upgraded AlexNet uses the Golden Jackal Optimisation Algorithm (GJOA) to determine the optimal feature extraction technique tuning after processing the input images. Moreover, the Fruit Shift Self-Attention Transform Mechanism (FSSATM) serves as the final classifier. This transform mechanism combines spatial position encoding (SPE) with a spatial feature extraction module (SFE) to increase the transformer's accuracy. 

Keywords: Golden Jackal Optimisation Algorithm; Fruit Shift Self Attention; Transform Mechanism; Modified AlexNet; Automated Fruit Identification; Spatial Feature; Extraction Module; Spatial Position Encoding.

Received on: 13/01/2025, Revised on: 23/03/2025, Accepted on: 06/05/2025, Published on: 22/11/2025

DOI: 10.69888/FTSCL.2025.000486

FMDB Transactions on Sustainable Computer Letters, 2025 Vol. 3 No. 4, Pages: 198-212

• Views : 11
• Downloads : 3

Download PDF