New species orchid recognition system using convolutional neural network / Annisa Atikah Mohd Fadzil and Itaza Afiani Mohtar

• Main Authors: Mohd Fadzil, Annisa Atikah (Author), Mohtar, Itaza Afiani (Author)
• Format: Book
• Published: Universiti Teknologi MARA, Perak, 2021-11.
• Subjects: Article; PeerReviewed
• Online Access: Link Metadata
• Notes: https://ir.uitm.edu.my/id/eprint/61546/1/61546.pdf; New species orchid recognition system using convolutional neural network / Annisa Atikah Mohd Fadzil and Itaza Afiani Mohtar. (2021) Mathematical Sciences and Informatics Journal (MIJ) <https://ir.uitm.edu.my/view/publication/Mathematical_Sciences_and_Informatics_Journal_=28MIJ=29/>, 2 (2). pp. 35-43. ISSN 2735-0703
• Keywords: QA Mathematics; Instruments and machines; Computer software; Neural networks (Computer science)

Orchid is famous for its variety and its beauty. Every year, about a hundred of new species names are published. The study is to determine whether the orchid species can be recognized using a convolutional neural network algorithm and to test the accuracy of the classification model. Transfer learning was also implemented in this project to skip the feature extraction phase that requires many computational resources in the CNN algorithm. The model of transfer learning that is used is the Inception V3 model. This project is to prove that the concept of new orchid species recognition can be done. The web application that created using HTML and Flask was able to recognize new species based on existing species. In this project, 10 existing species with 100 images each was selected in training, validating, and testing phase. The training accuracy reached 97% and the functional testing of orchid recognition results shows 83% accuracy with 1000 datasets. In conclusion, the use of a web system as a prototype tool for the recognition of new orchid species is helpful for the unlicensed persons/organization.