Análisis comparativo de ML Kit y TensorFlow Lite para el reconocimiento facial en aplicaciones multiplataforma

Abstract

Abstract: The following work presents a comparative analysis between Google ML Kit and TensorFlow Lite applied to face detection in cross platform mobile applications developed with Flutter. This study aims to contribute with an experimental analysis that facilities informed decision-making when selecting machine learning tools, given limited comparative documentation available for these two google solutions. The analysis involves the development of two independent prototypes, one using Google ML Kit and the other using the face_detection_full_range.tflite model from TensorFlow Lite, evaluated using the WiderFace Benchmark Dataset. Metrics such as precision, recall, F1 score, and processing time were measured, highlighting differences in performance, strengths, and limitations between the models. The applied methodology is experimental, ensuring objective and replicable results. The findings provide a clear understanding of the advantages and disadvantages of both tools.