There are numerous software applications, apps, that can identify birds, trees, flowers, and other living things all with varying degrees of accuracy. New research published in the International Journal of Intelligent Engineering Informatics offers a new approach to flower identification.
Abdulrahman Alkhonin, Abdulelah Almutairi, Abdulmajeed Alburaidi, and Abdul Khader Jilani Saudagar of the Information Systems Department at the Imam Mohammad Ibn Saud Islamic University in Riyadh, Saudi Arabia, explain how flowers are a big part of our lives in the aesthetic and recreational, educational, and even medicinal contexts and beyond.
Deep learning algorithms have been widely used recently in the fields of image processing and computer vision.
The team’s new algorithm has been trained on a variety of photos of four well-known flower types – sunflower, dandelion, rose, and tulip. The resulting application tested with colour photos on the Android mobile operating system could then identify new photos of dandelion flowers with an accuracy of 94.6%, sunflowers at 92.5%, tulips with 95.7%. For roses the recognition rate was a little lower at just under 90%. They explain that increasing the training data set will allow the accuracy of the algorithm to be improved.
The team adds that in future work they will incorporate an augmented reality feature in the application as extension that would help with flower identification out in the field, as it were.
Alkhonin, A., Almutairi, A., Alburaidi, A. and Saudagar, A.K.J. (2020) ‘Recognition of flowers using convolutional neural networks’, Int. J. Intelligent Engineering Informatics, Vol. 8, No. 3, pp.186–197.