A group of students, affiliated with the Graduate School of Artificial Intelligence and the Department of Industrial Engineering at UNIST distinguished themselves at the 2020 LG AI Hackathon.
The team Lab Fairy, made up of Yejin Kim, Jaeho Kim, Hongjun Yang, and Dongcheol Lim, won the third runner-up award in the 2020 LG AI Hackathon: Predicting SMILES for Molecular Structure Images, which took place from Tuesday, September 1 to Friday, October 9, 2020. This event has been organized by LG Sciencepark and hosted by Daycon Inc.
The challenge in this competition concerned AI for machine and computer vision, which required each team to develop AI-based algorithm system, using chemical structure images available, and translate those images into their corresponding SMILES strings.
SMILES (Simplified Molecular Input Line Entry System) is a chmical notation that allows users to represent a chemical structure in a way that can be used by the computer. The goal of this competition is to develop an AI model that can easily identify and expres the images of molecular structure in documents, such as papers and patents.
The four students are affiliated with Professor Junghye Lee’s Data Mining Lab and Professor Sungbin Lim’s Learning Intelligent Machine Lab. Prior to the competition, the team went through the process of data generation and sampling, data preprocessing, model training, ensemble learning, and GPU and multiprocessing. Upon submission, the team continued to stay on top, yet ended up finishing in third place.
Each winning team has been presented with an award certificate in addition to prize money. The first place winner was given KRW 5 million. The second and third placed teams of the hackathon were given KRW 3 million and KRW 2 million, respectively. One noticeable privilege of this competition is that winners with outstanding performance will be given the internship opportunity at the LG Sciencepark, and thus will be also exempted from the document-screening process.
Due to the outbreak of COVID-19, the competition was held in a non-contact manner and each team was permitted to have a maximum of five people. The award ceremony was held on Thursday, October 29, 2020.