Tsubasa Hirakawa

Chubu University

Papers

13

Total Citations

150

H-Index

6

About

Tsubasa Hirakawa is a computer vision and robotics researcher whose work sits at the intersection of deep learning, visual explanation, and intelligent manipulation systems. His research primarily advances object detection, semantic segmentation, and robotic grasping, with a particular focus on making neural network decision-making interpretable and practically deployable. Hirakawa is perhaps best known for the MT-DSSD framework, a multi-task deconvolutional single shot detector that simultaneously performs object detection, semantic segmentation, and grasp-point estimation within a unified network — a significant contribution to logistics robotics that has accumulated 50 citations across its iterations. Equally influential is his Attention Branch Network, which bridges visual explanation and performance improvement by leveraging attention mechanisms to both interpret and enhance CNN predictions, garnering 34 citations since 2018. His commitment to explainability extends into deep reinforcement learning, where he has developed attention-based methods for visualizing agent decision-making in actor-critic architectures, addressing the critical "black-box" problem in autonomous systems. Additional contributions span natural language generation for domestic service robots and automated data collection pipelines that reduce costly manual labeling. Collectively, Hirakawa's body of work reflects a sustained effort to build robotic systems that are not only capable, but transparent and practically scalable.

Research Focus

Key Achievements

6
H-Index
13
Papers
150
Total Citations
12
Avg Citations/Paper
🏆 Most Cited Paper
MT-DSSD: Deconvolutional Single Shot Detector Using Multi Task Learning for Object Detection, Segmentation, and Grasping Detection
34 citations · 2020
📈 Most Prolific Year: 2020 (3 Papers)
🤝 Key Collaborators: 22
🏛 Institutions: Chubu University

Top Papers

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10

Key Collaborators

Contact & Links

Available for collaboration
Content generated · 14 days ago