Micro-Video Understanding with Noisy Labels

Micro-videos are taking the social media world by storm with the rising of some online services, like Vine and Kwai. They benefit lots of commercial applications, such as brand building. Despite their value, micro-videos analysis is non-trival: 1) micro-videos are short in length and of low quality; 2) they can be described by multiple heterogeneous channels; 3) they are organized into a hierarchical ontology in terms of semantic venues; and 4) there are no available benchmark dataset on micro-videos. This project aims to build state-of-the-art deep learning models to understand the content in the micro-videos with noisy labels. The project expects to produce the next great step for machine intelligence-the potential to explore noisy labels to better understand, interpret and infer the content. This should benefit science, society, and the economy nationally through the applications of sensor technologies and cybersecurity.