Proceedings of the First International Conference on Combinatorial and Optimization, ICCAP 2021, December 7-8 2021, Chennai, India

Research Article

Deep Learning Architecture For Fruit Classification

Download126 downloads
  • @INPROCEEDINGS{10.4108/eai.7-12-2021.2314483,
        author={Sangeetha  B and Senthil Prabha  R and Ravitha Rajalakshmi  N},
        title={Deep Learning Architecture For Fruit Classification},
        proceedings={Proceedings of the First International Conference on Combinatorial and Optimization, ICCAP 2021, December 7-8 2021, Chennai, India},
        publisher={EAI},
        proceedings_a={ICCAP},
        year={2021},
        month={12},
        keywords={image recognition; deep learning; convolutional neural network; fruit recognition system},
        doi={10.4108/eai.7-12-2021.2314483}
    }
    
  • Sangeetha B
    Senthil Prabha R
    Ravitha Rajalakshmi N
    Year: 2021
    Deep Learning Architecture For Fruit Classification
    ICCAP
    EAI
    DOI: 10.4108/eai.7-12-2021.2314483
Sangeetha B1,*, Senthil Prabha R1, Ravitha Rajalakshmi N1
  • 1: PSG College of Technology
*Contact email: bsg.it@psgtech.ac.in

Abstract

The agricultural industries are one of the cost demanding fields placing a requirement on skilled laborers for harvesting. To meet the demands, robots are employed to harvest which mandates the need for accurate fruit detection system. The robot has to scan the image and recognize the fruit, which is the crucial process as the recognition system faces unprecedented challenges like occlusion, deformation, illumination conditions. The objective of this work is to build an accurate and reliable fruit recognition system by addressing these challenges in image recognition. Convolutional neural network, a deep learning algorithm is used to identify the features of an image and classify the image in the fruit recognition system. The system is evaluated with Fruit-360 dataset consisting of 43329 images of 60 different categories. With the aid of the proposed system, quantifiable improvement of about 97% accuracy is achieved and the total loss of the system is about 0.13.