🎓 Biography

I am a Machine Learning Engineer at Woven by Toyota, building production perception systems for Woven City’s Digital Twin Platform. My work spans 3D reconstruction, multi-camera tracking, and vision-language reasoning, with GPU training and inference pipelines on AWS serving teams in Japan and overseas. I hold a Ph.D. in Engineering from the Institute of Science Tokyo. My research focuses on 3D vision, autonomous driving perception, and generative modeling. I review for NeurIPS, ICLR, ICML, and CVPR.

💼 Experience

• Woven by Toyota, Machine Learning Engineer
  Tokyo, Japan · Aug 2025–Present
  Build the multimodal perception pipeline for Digital Genba’s logistics digital twin, from data capture and 3D Gaussian Splatting to multi-camera detection and tracking and VLM-based behavior analysis and reasoning. Run production GPU workflows on AWS and deliver reusable ML capabilities to customer teams in Japan and overseas.

• Sony, R&D Intern
  Tokyo, Japan · Jan 15–Mar 31, 2025
  Trained a multi-view diffusion model on the GAIA GPU cluster to generate synthetic street-view and bird’s-eye-view data for autonomous perception. Built a pipeline to synchronize LiDAR, images, 3D boxes, and text.

• Preferred Networks, R&D Intern
  Tokyo, Japan · Aug 5–Sep 30, 2024
  Developed a controllable diffusion model for background generation and tested its generalization against established methods across datasets and artistic foreground styles.

• SenseTime Japan, Researcher Intern
  Tokyo, Japan · Jun 5–Nov 30, 2023
  Co-developed VSRD (CVPR 2024), generating 3D pseudo-labels from 2D masks to eliminate manual 3D box annotation for training. On KITTI-360 Hard, VSRD raised BEV AP@0.3 from 37.34 to 41.28 versus Autolabels with the same detector.

• MEGVII, Computer Vision Engineering Intern
  Shanghai, China · Jan 5–Mar 30, 2022
  Improved parking-slot detection recall by 10% using geometric constraints and targeted data augmentation for occlusion and image-edge cases. MEGVII adopted the model in its automatic-parking perception pipeline.

• Tokyo Institute of Technology, Research Assistant
  Tokyo, Japan · Apr 1, 2024–Present
  Collaborated with a Japanese mapping company on large-scale high-definition (HD) map reconstruction. Investigated 3D Gaussian Splatting (3DGS) for rendering large-scale scenes.

🔔 News

• [06/2026] Our paper StereoSplat+ was accepted at IROS 2026. The arXiv preprint is available.
• [06/2026] Received Best Paper Award, Symposium on Sensing via Image Information (SSII 2025).
• [03/2026] I have received the “Graduate Research Encouragement Award” from the Society of Automotive Engineers of Japan (JSAE).
• [01/2026] I have completed my PhD defense presentation for the doctoral thesis in the Institute of Science, Tokyo, titled “Deep Learning Based Stereo Vision: From Robust Stereo Matching to Feed-Forward 3D Reconstruction.”
• [07/2025] Our paper DMS was accepted for an oral presentation at the ICCV 2025 AIM Workshop.
• [01/2025] Our paper titled “TDM: Temporally-Consistent Diffusion Model for All-in-One Real-World Video Restoration” has been selected as the Best Student Paper for MMM2025.
• [10/2024] One paper was accepted by MMM 2025.
• [07/2024] One paper was accepted by BMVC 2024.
• [02/2024] One paper was accepted by CVPR 2024.
• [02/2024] One paper was accepted by ICRA 2024.
• [10/2023] One paper was accepted by WACV 2024.
• [06/2023] Started internship at SenseTime Japan.
• [04/2023] Received Tokyo Tech Advanced Human Resource Development Fellowship.
• [03/2023] Received the SCIE Student Award.
• [03/2023] Received Excellent Graduation Paper Award.
• [06/2022] One paper was accepted by ACM MM 2022.

📝 Publications

IROS 2026 StereoSplat+: Feed-Forward Stereo Gaussian Splatting with Diffusion-Assisted Progressive Inference

Zihua Liu, Masatoshi Okutomi
IEEE/RSJ International Conference on Intelligent Robots and Systems
📄 Paper     🌐 Project     💻 Code     ▶ Demo

ICCVW 2025 DMS: Diffusion-Based Multi-Baseline Stereo Generation for Improving Self-Supervised Depth Estimation

Zihua Liu, Yizhou Li, Songyan Zhang, Masatoshi Okutomi
ICCV 2025 AIM Workshop · Oral Presentation
📄 Paper     💻 Code     📎 Supplement

CVPR 2024 VSRD: Volumetric Silhouette Rendering for Weakly Supervised 3D Object Detection.

Zihua Liu*, Hiriki Sakuma, Masatoshi Okutomi
Proceedings of the IEEE/CVF Computer Vision and Pattern Recognition.
📄 Paper     🌐 Project     💻 Code

ICRA 2024 CFDNet: A Generalizable Foggy Stereo Matching Network with Contrastive Feature Distillation.

Zihua Liu, Yizhou Li, Masatoshi Okutomi
IEEE International Conference on Robotics and Automation
📄 Paper     🌐 Project     💻 Code

WACV 2024 Global Occlusion-Aware Transformer for Robust Stereo Matching.

Zihua Liu, Yizhou Li, Masatoshi Okutomi
Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision
📄 Paper     🌐 Project     💻 Code

ACM MM 2022 Digging Into Normal Incorporated Stereo Matching.

Zihua Liu, Songyan Zhang, Zhicheng Wang, Masatoshi Okutomi
Proceedings of the 30th ACM International Conference on Multimedia
📄 Paper     🌐 Project     💻 Code

BMVC 2024 Disparity Estimation Using a Quad-pixel Sensor.

Zhuofeng Wu, D.Lee, Zihua Liu, K.Yoshizaki, Masatoshi Okutomi
The Thirty Fifth British Machine Vision Conference
📄 Paper     🌐 Project     💻 Code     📦 Dataset

MMM 2025 🏆 Best Student Paper TDM: Temporally-Consistent Diffusion Model for All-in-One Real-World Video Restoration

Yizhou Li, Zihua Liu, Yusuke Monno, Masatoshi Okutomi
MultiMedia Modeling: 31st International Conference on Multimedia Modeling
📄 Paper     🌐 Project     💻 Code

🏅 Honors & Awards

• Tokyo Tech Advanced Human Resource Development Fellowship.      2023/04-2026/04
• SCIE Student Award.      2023/03

🤝 Services

• Reviewer for the CVPR2025, ICML2025 and ICLR2025      2025

• Reviewer for the ACM International Conference on Multimedia      2022,2024

• Reviewer for the BMVC     2024

• Reviewer for the NeurIPS     2024

• Reviewer for the ICLR     2024