| dc.description.abstract | The facial image of a person contains much rich information like gender, race, age, hairstyle, and other essential pieces of information that are related to a person�s occupation. Predicting or recognizing occupation is a classification problem. Because of its wide reach in intelligent systems and services, this is a feasible computer vision challenge. In this thesis, we have stated a related work that includes the different studies on this topic. The previous study was done based on recognizing occupation from human clothing, scene context, social context, and a facial image of a person. My thesis objective is to predict occupation from a facial image of a human. We have collected data on Indian people and made a new dataset. In this ork, we have used two different datasets. The first dataset, DB1, is used in previous studies as well, which is based on eastern Asian people. The second dataset is based on Indian people, which is made by us. Both the dataset contains five different classes. We have used multiple image classification algorithms and compared them in terms of accuracy and performance. Out of all of these algorithms that we have used, The vision transformer performs best as compared to other algorithms. Also, the vision transformer achieves better accuracy as compared to all other previous works, which is based on occupation prediction from the facial image. | |
