http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
마부불 General Graduate School of Korea Aerospace Univers 2022 국내박사
The lifestyle of humans has changed noticeably since the contagious COVID-19 disease struck globally. People should wear a face mask as a protective measure to curb the spread of a contagious disease. Therefore, real-world applications dealing with human ages extracted from face images must migrate to a robust system proficient to estimate the age of a person wearing a face mask. As a part of this thesis, a hierarchical approach for age estimation from masked facial images is proposed in a group-to-specific manner rather than a single regression model because age progression across different age groups is quite dissimilar. The intention was to squeeze the feature space among limited age classes so that the model could fairly discern age. In the same way, this pandemic changed the human shopping nature, necessitating a contactless shopping system to protect oneself as well as the community effectively. Thus, a customer opts for a store where it is possible to avoid physical contact and shorten the shopping process with extended services such as personalized product recommendations. Automatic age and gender estimation of a customer in a smart store strongly benefit the consumer by providing personalized advertisement and product recommendations; additionally, it aids the smart store proprietor to promote sales and develop an inventory perpetually for future retail. In this regard, a deep learning-founded enterprise solution is proposed for smart store customer relationship management (CRM), which allows us to predict the age and gender from a customer’s face image taken in an unconstrained environment to facilitate the smart store’s extended services, as it is expected for a modern venture. For the task of age estimation from a masked and non-masked facial image, the large public IMDB-WIKI dataset is used. In particular, a synthetic masked version of the IMDB-WIKI dataset is generated to train and validate the proposed hierarchical approach since there is no masked face image benchmark dataset with real-age annotations. The in-the-wild Adience dataset was utilized for training and validation of the network during the gender estimation task. The data sparsity problem of the large public IMDB-WIKI dataset is fairly mitigated using off-the-shelf down-sampling and up-sampling techniques as required. The age estimation task was fully modeled like a deep classification problem, and expected ages were formulated from SoftMax probabilities. The classification task for non-mask age and gender estimation is handled utilizing an empirically leading pre-trained convolutional neural network (CNN), the VGG-16 network, and incorporates batch normalization in the intermediate layers whereas multiple low-memory and higher-accuracy-based CNNs are deployed for masked face age estimation tasks. The systems are validated with two standard benchmarks IMDB-WIKI and Adience, one for each task, and demonstrate SOTA performance for both real age and gender estimation. Thus, the proposed hierarchical framework of the masked face age measurement system showed 0.86-year progress in respect of mean absolute error (MAE) compared with the singular model method. Moreover, this study is probably the first effort to measure a person's actual age from his or her masked facial image.