Improving Borderline Adulthood Facial Age Estimation through Ensemble Learning

摘要: Achieving high performance for facial age estimation with subjects in the borderline between adulthood and non-adulthood has always been a challenge. Several studies have used different approaches from of baby to an elder adult datasets employed measure mean absolute error (MAE) ranging 1.47 8 years. The weakness algorithms specifically motivation this paper. In our approach, we developed ensemble technique that improves accuracy underage conjunction deep learning model (DS13K) fine-tuned on Deep Expectation (DEX) model. We achieved 68% group 16 17 years old, which is 4 times better than DEX such range. also present evaluation existing cloud-based offline prediction services, as Amazon Rekognition, Microsoft Azure Cognitive Services, How-Old.net DEX.