As a result of learning facial images with EfficientNet based on CNN structure for Koreans, Chinese, and Japanese who have similar appearances among races sharing the same culture, we confirmed a high accuracy of about 92%. The high recognition rate of EfficientNet is encouraging, but more efficient methods are required as it takes a lot of time to build and process the model. Therefore, this paper aims to create an algorithm that can quickly analyze facial features that have the greatest influence on classifying race. To this end, first, the percentage of a specific area is calculated by using Face Landmark to dot the entire outline of the face and measuring each distance from the reference point. Second, based on features determined by the neural network through Grad-CAM among eXpliable AI (XAI) techniques, the ratio values of specific regions are learned by machine learning algorithms and DNN. As a result, it was confirmed that the method using Face Landmark was the highest with an accuracy of about 63%.
한국어
같은 문화를 공유하는 인종 간에 비슷한 외모를 가지고 있는 한국, 중국, 일본인에 대해 CNN 구조를 기 반한 EfficientNet으로 얼굴 이미지를 학습시킨 결과, 약 92%의 높은 정확도를 확인하였다. EfficientNet의 인식률 이 높다는 점은 고무적이지만, 모델을 빌드하고 처리하는 것에 많은 시간이 필요하므로 좀 더 효율적인 방법이 요 구된다. 따라서 본 논문은 인종을 구분하는 데에 가장 큰 영향을 미치는 얼굴의 특징을 알아내어 빠르게 분석할 수 있는 알고리즘을 만들고자 한다. 이를 위하여 첫째, Face Landmark를 사용하여 얼굴의 전체 윤곽을 점으로 표시하고 기준점으로부터의 각 거리를 측정하여 특정 부위의 비율을 계산한다. 둘째, eXplainable AI (XAI) 기법 중 Grad-CAM을 통해 신경망이 판단한 특징을 기반으로 특정 영역의 비율 값을 머신러닝 알고리즘과 DNN으로 학습한다. 그 결과 Face Landmark를 활용한 방법이 약 63%의 정확도로 가장 높은 것을 확인하였다.
목차
요약 Abstract Ⅰ. 서론 Ⅱ. 데이터 전처리 및 학습환경 Ⅲ. 학습 모델 구성 및 결과 Ⅳ. 모델 특성 탐색 Ⅴ. 결론 REFERENCES
Ever since next generation convergence technology became one of the most important industries in the nation, computing professionals have encountered a growing number of challenges. Along with scholars and colleagues in related fields, they have gathered in avariety of forums and meetings over the last few decades to share their knowledge, experiences and the outcome of their research. These exchanges have led to the founding of the International Next-generation Convergence technology (INCA) on December 1, 2015. INCA was registered as an incorporated association under the Ministry of Information and Communications. The main purpose of the organization is to improve our society by achieving the highest capability possible in next generation convergence technology.
간행물
간행물명
차세대융합기술학회논문지 [The Journal of Next-generation Convergence Technology Association]