Abstract—Facial landmarks are places on the human face that are used to explain morphology. Facial landmarks are employed in anthropology to minimize the quantity of data utilized in morphological analysis, and they have become one of the pillars of anthropological study over the years. Their applications in facial analysis range from identifying morphological changes throughout facial development to identifying facial similarities. Despite its extensive use, most of the research in automatic facial landmark recognition has been motivated by the necessity for automation of security activities such as face identification and verification. Approaches using 2D pictures models have been widely researched in order to offer reliable face landmark recognition in a variety of real-world scenarios. Many current techniques, for example, rely on landmark recognition on low resolution data with significant out-of-plane rotation1 and occlusions. This is in sharp contrast to the data utilized in anthropology, which is comprised of high quality 2D photometry. The data is frequently collected from consenting participants and adjusted to minimize non-shape facial changes, such as varying size owing to the subject's distance from the capturing equipment or head rotation. The datasets in anthropology purpose are unavailable in comparison to those used for training current state-of-the-art facial landmark detection approaches based on Improved Faster Region Convolutional Neural Networks (Faster R-CNN). Therefore, an improved in data of anthropometric images by augmentation and change Roi Pooling by Roi Align that could promise an optimized and enhance the accurate with less time.
Index Terms—faster R-CNN, small facial landmark, Vietnamese face, landmarks detector, anthropology
Cite: Ho Nguyen Anh Tuan, Nguyen Dao Xuan Hai, and Nguyen Truong Thinh, "The Improved Faster R-CNN for Detecting Small Facial Landmarks on Vietnamese Human Face Based on Clinical Diagnosis," Journal of Image and Graphics, Vol. 10, No. 2, pp. 76-81, June 2022.
Copyright © 2022 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.
Copyright © 2012-2023 Journal of Image and Graphics, All Rights Reserved