Face image mining using microsoft api service

• 主要作者: Tan, Aloysius Han Tian.
• 其他作者: Teoh Eam Khwang
• 格式: Final Year Project
• 語言: English
• 出版: 2011
• 主題: DRNTU::Engineering
• 在線閱讀: http://hdl.handle.net/10356/44340
• 機構: Nanyang Technological University

Human faces do convey a significant amount of information and contained important attributes to perform age related applications. The lack of good public aging face dataset restricts the work in research and development works. In this project, the author presents a new automated, innovative and robust face image mining application and using API offered by Microsoft Bing Service. The images of interest can be retrieved from the Internet with the different key modules developed by the author. The main contribution towards this project was to develop modules that seek to perform their designed function and implement innovative methods in order for the new application to extensively and accuracy mine for all human face images of all age group over the World Wide Web. The Image Gathering Modules provides features such as an automated mechanism to perform different age related search queries that are pre-constructed in the module to search for all human ages using Bing API. A filtering mechanism was implemented to extensively filter noisy images and to acquire the desirable image information. An Image Extraction Module performs an automation extraction of all relevant images from the Internet, checking for duplication of images and with an additional layer of noise filtering. Lastly a validation module was created for human validation. The user interface created allows users to do selections or update image information based on the different category. A clean human face dataset of all ages can be obtained and be used for age estimation after validation. The application achieved the objective to intensively mine for a large human aging facial dataset with a total of 101k images were collected. This figure can be increase significantly by constructing more search query per age. To increase the accuracy of the relevant images dataset collected, the module can pre-construct a more age specific related search query when perform mining for human face images using the application.