Tag: kinship verification competition

Computer science students participated in the IEEE International Conference on Automatic Face and Gesture Recognition 2021 (FG 2021) Kinship Verification challenge as part of their Deep Learning (DL) course, taught by adjunct Associate Professor Iddo Drori. They presented their findings at the conference. 

The goal of the kinship verification competition is to determine whether a parent-child, sibling, or grandparent-grandchild relationship exists between two people. It is important in social media applications, forensic investigations, finding missing children, and reuniting families. The team demonstrated high-quality kinship verification by participating in the FG 2021 Recognizing Families in the Wild Challenge which provides the largest publicly available dataset in the field. Their approach, winning third place in the competition, ensembled models written both by humans and written automatically by OpenAI Codex for the first time.

As part of the unique Deep Learning course curriculum, students get to compete in common task framework competitions, which enable them to test the waters in the real-world while advancing science. This semester Drori and teaching assistants Newman Cheng and Vaibhav Bagri performed feasibility tests on the Kinship Verification Challenge and found it in line with the course goals.

Students used the kinship verification dataset to design, develop, and train deep learning models. Over the course of three weeks, teams worked on and improved their submissions competing with groups from across the world. Drori then worked with the leading teams by using OpenAI Codex to improve the verification models even further. The effort paid off with the students at the top three of the competition leaderboard, claiming victory by quickly writing up their findings.

The winning team is composed of graduate students Junyi Huang (Mathematics), Maxwell Strome (Computer Science), Ian Jenkins (Applied Physics and Math), Parker Williams (Computer Science), Bo Feng (Electrical Engineering), Yaning Wang (Electrical Engineering), and Roman Wang (Computer Science).

“Winning third place in this international challenge is an excellent achievement. The teams used both humans and machines to automatically write the code, which is a first and commendable feat!” said Drori.