Fruit Detection and Recognition Using Faster R-CNN with FPN30 Pre-trained Network

• Published in: 8th International Conference on Recent Advances and Innovations in Engineering: Empowering Computing, Analytics, and Engineering Through Digital Innovation, ICRAIE 2023
• Main Author: Roslan M.I.; Ibrahim Z.; Adnan N.A.K.; Diah N.M.; Narawi N.A.A.; Arif Y.M.
• Format: Conference paper
• Language: English
• Published: Institute of Electrical and Electronics Engineers Inc. 2023
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85189941080&doi=10.1109%2fICRAIE59459.2023.10468169&partnerID=40&md5=1508755d6c78027cd39afff22ecc3388

Accurate and reliable fruit detection and recognition in orchards is critical for enabling higher-level agriculture tasks such as fruit picking. However, detecting and recognizing fruits with occlusion by neighboring fruits is extremely difficult. Faster R-CNN (Faster Region-based Convolutional Neural Network) is a well-known deep learning technology for object detection and recognition. Thus, this study investigates the application of Faster R-CNN for apple detection and recognition. Two different datasets have been constructed under variable illumination conditions and occlusion; an inter-class dataset that consists of images of apples and oranges, and an intra-class dataset that consists of images of two types of apples, namely fuji and royal gala apples. Results indicate that Faster R-CNN can detect and recognize apples from oranges, and the fuji apple in the orchards, with high accuracy. This suggests that Faster R-CNN can be used practically in the real orchard context. © 2023 IEEE.