ELAION: ML-Based System for Olive Classification with Edge Devices

Abstract

Discrimination of morphological characteristics for olive fruits is widely used to quickly classify their cultivar. This process is usually based on visual observations, which require experience and often appear to be very subjective, inconsistent and inaccurate. Towards automating and providing an error-free procedure for olive fruit classification, this work presents ELAION, an end-to-end system for olive cultivar identification using edge devices, such as smartphones and tablets. An application utilizes the device’s camera to send olive images to a back-end server for feature extraction. Results are relayed back to the application, which identifies the originally depicted olive cultivar using pre-trained machine learning models. As a result, ELAION greatly reduces the time and errors on olive fruit identification with on-site results, thus paving the way for becoming an on-site key-tool for olive growers, breeders, and scientists.

This work (T2E$\Delta$K-02637) was co-financed by the Special Managing and Implementation Service in the areas of Research, Technological Development and Innovation (RTDI) - Greece, and the European Union.