2026-10-09 Hot Topic

A paper by Masato Taki, Team Director of the AI as Science Team at RIKEN iTHEMS and Associate Professor at Rikkyo University, together with Soshun Kihara and Shunsuke Yasuki, graduates of Rikkyo University's Graduate School of Artificial Intelligence and Science, has been accepted at NeurIPS 2026, a leading international conference in AI and machine learning.

The researchers used knowledge distillation to transfer what an AI model learned from event-camera data, which capture changes in brightness at individual pixels, to a model processing conventional color images. The resulting model relied more on object shapes than on color or texture and became less sensitive to color changes and fine-grained noise.

For details, please see Rikkyo University's press release via the related link.

Related Link