Application for Image Processing in a Receiver Operating Character-Based Testing Review

The image processing technology is currently developing quite rapidly. Implementation of technology based on artificial intelligence systems has entered various sectors and platforms, mostly in the health and agriculture sectors. There are challenges in developing image processing applications, especially on mobile platforms, including processing a large amount of data that involves millions of pixels that must be recognized. To measure the quality of the results of the process certainly requires a reliable framework. In testing the quality of training data in the database and case studies, many image processing techniques utilize the Receiver Operating Character (ROC) approach’s reliability. In some cases, this ROC approach is very different depending on the data model, output, and algorithm used. This study aims to review ROC’s use as an approach used in testing systems on developed mobile applications. The case studies used in this research are image processing in agriculture cases for pests and disease recognition in cocoa fruits and chili leaves, which is then implemented in a mobile application. Testing entities in the ROC approach are analyzed based on actual and predicted conditions. The measurement results show that the ROC approach is applied in both test cases. It can only produce accuracy measurements on the value of accuracy, sensitivity, and precision. The values on chili leaves are accuracy 60%, sensitivity 60%, and precision 100%. The accuracy is 66.7% for cacao fruit objects, sensitivity 66.7%, and precision 100%.