SBIR Phase I: Face Analyzer / Semantic Search

点击翻译按钮获取中文摘要

The broader/commercial impact of this Small Business Innovation Research (SBIR) Phase I project is significant as the company’s advanced Face Analysis AI technology will accelerate AI projects by 18-24 months. This innovative technology is poised to have a positive influence on various sectors such as retail and public safety, offering applications that go beyond facial recognition/identification. By focusing on collaborative human-AI facial tracking and analysis, this technology addresses ethical concerns and mitigates the risks and consequences of traditional facial recognition technologies. The project promotes diversity and inclusion in STEM fields by emphasizing a diverse team to combat bias and equity issues in technology development. The technology can contribute to national defense efforts by enabling efficient search for facial attributes of interest based on semantic queries. This aspect is particularly relevant in public safety scenarios with large crowds and high-security concerns. By reducing bias, improving accuracy, and addressing privacy and ethical concerns, the technology can have a lasting impact on the AI industry while advancing the welfare of the American public and supporting security efforts.This Small Business Innovation Research (SBIR) Phase I project aims to create Face Analyzer/Semantic Search, an AI system bridging descriptive text and facial photos. Unlike conventional face recognition systems, which necessitate a probe photo for comparisons, the company's innovation seeks to eliminate this requirement. This approach offers benefits in terms of time, cost, and accuracy, challenging the conventional wisdom in the field. The project's initial challenge involves assembling diverse training datasets with labeled face photos and textual descriptions, establishing a scalable data pipeline to enhance accuracy and mitigate bias. The second challenge is assessing the accuracy of facial attribute classification models derived from text and images across various attributes, image types, sizes, and ambient conditions. The third challenge involves optimizing model size and computational efficiency for cost-effective deployment. The proposed solution entails constructing a comprehensive training image dataset, expanding computer vision capabilities, developing a natural language processing module, and implementing a matching system. Key milestones for product development include creating precise facial image indexing modules, enabling the extraction of facial attributes from textual descriptions, and efficiently deploying the system in the cloud. The innovation promises to streamline and enhance facial analysis, potentially reshaping the field of face AI.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.