Socar Hosts World's First Mobility Mega AI Challenge

Socar is conducting a mobility super-large AI challenge involving AI researchers from around the world. The photo shows the Socar AI team. (Photo by Socar)

Socar is conducting a mobility super-large AI challenge involving AI researchers from around the world.

On the 19th, Socar announced that it will host the workshop ‘VPLOW (Visual Perception via Learning in an Open World)’ at 'CVPR 2023,' the world's most prestigious conference in the field of computer vision held in June, and hold the first-ever super-large AI global challenge in the mobility domain.

The Socar AI team is hosting the ‘FMDC (Foundational Model without Descriptive Caption)’ challenge with the theme of ‘multimodal-based few-shot learning’ until June 15. Participants will undertake the task of building multimodal AI models that understand and learn data expressed in different formats such as text and images like humans, and compete in the challenge. The final winning team will have the opportunity to present their results at the CVPR 2023 workshop held in Vancouver, Canada, on June 18.

For this challenge, Socar provides participants free access to a dataset of over 30,000 image-text pairs actually used in Socar’s super-large AI. The dataset consists of images and texts acquired during the operation of the car-sharing platform, including indoor and outdoor photos of Socar vehicles that can identify damage or contamination. Personal information has been removed from the data, and AI ethics verification has been completed to prevent bias.

In the challenge, participants’ AI models are evaluated on how accurately they understand new data occurring in real life based on the data they have learned. For example, it is a method where a person who has never seen a damaged vehicle is shown a photo of a damaged vehicle along with a caption such as ‘There is a scratch on the left door of the vehicle’ simultaneously, learns the commonalities and differences of various objects, and then identifies the occurrence of an accident from images of accidents occurring in reality.

The Socar AI team said, “We are very honored to hold a challenge on AI technology, a core competency of operating a mobility platform, at the world’s most prestigious conference,” and added, “We hope this workshop and challenge will be a great opportunity to introduce Socar’s AI technology, which operates a large fleet of about 20,000 vehicles annually, to AI researchers worldwide and to jointly consider the development of the mobility technology ecosystem.”

This content was produced with the assistance of AI translation services.