Cao-nalyzer: An Android-Based Mold Detection in Cacao Beans Using Faster R-CNN Algorithm

Abstract

Mold-infested cacao beans present a substantial hurdle for the cacao industry, affecting not only the quality of the beans but also the cocoa products derived from them. With this issue, the proponents of the study developed an application aimed to detect moldy cacao beans using Faster Region-based Convolutional Neural Network algorithm for online mode and You Only Look Once version 8 for offline mode. After training the model, it resulted in a Mean Average Precision (mAP) of 83.96% in the Faster R-CNN algorithm and 98.16% in the YOLO algorithm. During the testing phase, the researchers conducted several tests such as Android Testing, Camera Resolution Testing, Ram Testing, Capture Distance Testing, Light Testing, Function Analysis Test, and Module Testing. The application then underwent refinements and enhancements based on their feedback.