Qiaomu Miao
About Me
Hi! I am Qiaomu Miao. I received my Ph.D. in Computer Science from Stony Brook University, advised by Prof. Dimitris Samaras and Prof. Minh Hoai Nguyen. My research expertise spans computer vision, vision-language models, multi-view geometry, generative AI, and human behavior analysis. Before that, I obtained my Bachelor’s and Master’s degrees in Computer Science at Tianjin University. During my Master’s study, I investigated the cognitive neural mechanisms of human vision.
Contact: qiamiao AT cs.stonybrook.edu
Publications
(2026).
OmniGF: A Dual-Branch Vision-Language Framework for Unified Gaze Following.
Conference on Neural Information Processing Systems (NeurIPS), 2026 (Spotlight).
(2026).
Phase-Aligned RoPE for Mixed-Resolution Diffusion Transformer.
European Conference on Computer Vision (ECCV), 2026.
(2026).
Behavior-Based Skill Assessment for Open Surgery from Multi-View and Egocentric Videos.
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2026.
(2025).
Multi-view Gaze Target Estimation.
International Conference on Computer Vision (ICCV), 2025.
(2024).
Diffusion-Refined VQA Annotations for Semi-Supervised Gaze Following.
European Conference on Computer Vision (ECCV), 2024.
(2023).
Patch-level gaze distribution prediction for gaze following.
IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2023.
(2022).
Study of detecting behavioral signatures within DeepFake videos.
Arxiv Preprint, 2022.
Projects
Experience
Research Scientist Jul 2026 - Present
Meta
- Working on content representation learning with multimodal input for large-scale autoregressive recommendation systems on Facebook Feed/Reels.
Student Researcher Sep 2025 - Apr 2026
- Developed eye segmentation and eye-tracking pipelines with large foundation models.
Technology Investigation Intern May 2022 - Aug 2022
Apple
- Designed and trained activity recognition models on video datasets for augmented reality (AR) applications
- Improved activity recognition performance by incorporating scene understanding concepts
Research Intern Jun 2021 - Aug 2021
Bytedance
- Generated motion-transferred DeepFake videos with state-of-the-art lipsyncing and face reenactment models to investigate the behavioral signatures (speaking style and utterance, etc.) in person identification
- Implemented a self-supervised DeepFake Detection model using behavioral and appearance-related features