Handling large amounts of data is essential when considering large-scale image processing. For such large-scale processing, we propose data structures and algorithms for fundamental technologies such as nearest neighbor search and clustering. We cover a wide range of topics, from low-level processing, such as SIMD acceleration, to high-level applications, such as providing the visual system as an open source library.

We are researching to improve the performance of basic data structures such as hash tables using machine learning techniques. By considering data distribution, we design systems with better performance trade-offs than traditional ones.

We pursue research in multimodal information processing, such as image and language. Using prompt engineering techniques, we seek new applications in multimodal spaces.