Koshiro Nagano

Hi, I’m Koshiro Nagano, a Ph.D. student at Keio University advised by Prof. Hideo Saito. I work as a Computer Vision Engineer at Konica Minolta, Japan. My research focuses on data-efficient deep learning, with particular interest in few-shot object detection and learning from synthetic data.

📝 Research Interests

• Sim2Real: Synthetic-to-Real Transfer Learning, Domain Adaptation

• Continual Learning: Streaming Video Understanding, Catastrophic Forgetting

• Edge AI: Efficient Learning Design, On-Device Inference

📰 News

• [2026/04] Our paper has been accepted to CVPR 2026 (Acceptance rate: ~25%)

• [2025/09] Our paper has been accepted to ICIP 2025

• [2023/12] I opened this site

📄 Selected Publications

Learning from Synthetic Data via Provenance-Based Input Gradient Guidance

Koshiro Nagano, Ryo Fujii, Ryo Hachiuma, Fumiaki Sato, Taiki Sekii, Hideo Saito

CVPR 2026, Acceptance Rate: 25.42%, h5-index: 450, 2026

Paper

Frozen Network Few-Shot Object Detection

Koshiro Nagano, Fumiaki Sato, Ryo Hachiuma, Kazuki Tsutsukawa, Taiki Sekii

ICIP 2025, 2025

Paper