Team of University of Miami Graduate Students Finishes Third Place in 2018 TRECVID Ad-hoc Video Search Competition
A team of University of Miami graduate students competed in the 2018 Text REtrieval Conference’s Video Retrieval Evaluation (TRECVID), finishing in third place. The students, Saad Sadiq, ECE PhD ’19; Yudong Tao, ECE PhD ’22; Yuexuan Tu, ECE PhD ’22; and Yilin Yan, ECE PhD ’18 were led by Mei-Ling Shyu, professor in the College of Engineering’s Department of Electrical and Computer Engineering.
The team participated in TRECVID’s Ad-hoc Video Search (AVS), where they were tasked with building a video retrieval system within a one-month period, capable of searching for 30 types of videos in a large-scale database. For each topic, the video retrieval system had to return all the video segments which met the search criteria, ranked in order of confidence.
“TRECVID’s competitions are very important because they encourage and stimulate progress in content-based video retrieval,” explains Shyu. “The advancements made from these competitions are especially important to video-sharing websites. These websites currently use metadata – a set of data that describes and gives information about other data – such as titles, thumbnails, and descriptions to return results from a search, which is sometimes inefficient and inaccurate.”
“The development of our system potentially boosts a search engine’s ability to understand text from a search query and relate that to a video’s content, without relying on descriptions of the video created by the uploader or owner,” Shyu continues.
The TRECVID is a conference series sponsored by the National Institute of Standards and Technology (NIST) that encourages and promotes research in content-based video retrieval and analysis via open, metrics-based evaluation. TRECVID functions as a laboratory-style evaluation that attempts to model real-world situations.