We will present our work on aligning Human-AI intentions via attention heads perception matching at PixFoundation Workshop of CVPR 2025. Please come have a look at our poster on Thursday. Paper link: openaccess.thecvf.com/content/CVPR...

Optimal Design of Wet Etching Bath for 3D Flash Memories Using Multi-Objective Bayesian Optimization Recently, the complexity of semiconductor manufacturing processes has increased, resulting in a growing need for high-precision optimization of device structures. For example, in batch-type wet etchin...

Our latest work on injecting user's visual preference into Bayesian Optimization loop has been accepted by IEEE Transactions on Semiconductor Manufacturing. ieeexplore.ieee.org/document/110...

We describe that in the era of vision foundation models, relevance feedback is a surprisingly effective technique in improving the performance of many different tasks in image retrieval/search, reflecting user preferences at the same time.

Image Retrieval Reflecting User Preferences | KIOXIA - Japan (English) We make foundation-model-based image retrieval able to handle various user’s preferences by leveraging relevance feedback. The proposed approach is empirically effective in various image retrieval scenarios.

We put up a technical blog post about our AI research work (ECCV 2024 Workshops) in collaboration with The University of Tokyo. www.kioxia.com/en-jp/rd/tec...

Our ACCV paper will be at the poster session on the final day of #ACCV2024. My colleague will be there. Come to say hi. Our paper is about attenuating artifacts in Vision Transformers. openaccess.thecvf.com/content/ACCV...