Tianyi Gao

Hi! I am a first-year CS PhD student in the Multimodal Vision Research Laboratory at Washington University (WashU), advised by Dr. Nathan Jacobs. I work on computer vision, multimodal learning and geospatial AI.

Before joining WashU, I obtained my Bachelor's and Master's degree from Wuhan University (WHU). I also spent a wonderful time at Microsoft Research Asia in 2024, working on MLLMs for geoscience.

Email / CV / Google Scholar / Github

Research

I am excited about building AI that understands humans and the environment through the lens of geospatial data, creating meaningful social impact. Through this process, I continue to learn from the gaps revealed and explore ways to improve current computer vision and multimodal learning techniques.

Currently, I am interested in equipping embodied agents with the capability to perceive and navigate urban environments, driven by curiosity about a future of human–robot coexistence: how they sense cities, interpret their pulse, and engage with human life.

	PEACE: Empowering Geologic Map Holistic Understanding with MLLMs Yangyu Huang, Tianyi Gao, Haoran Xu, Qihao Zhao, Yang Song, Zhipeng Gui, Tengchao Lv, Lei Cui, Scarlett Li, Furu Wei IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025 Microsoft Foundry Labs Project We introduce GeoMap-Bench, a vision-language benchmark for geologic map understanding. It consists of 25 task types, which measure abilities across 5 aspects. Our benchmark reveals a significant performance gap between state-of-the-art MLLMs and human experts, we further explore agentic baselines to improve the performance.
	Enrich Distill and Fuse: Generalized Few-Shot Semantic Segmentation in Remote Sensing Leveraging Foundation Model’s Assistance Tianyi Gao, Wei Ao, Xing-Ao Wang, Yuanhao Zhao, Ping Ma, Mengjie Xie, Hang Fu, Jinchang Ren, Zhi Gao IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW, Oral), 2024 We incorporate general VLMs into existing GFSS pipelines through support set augmentation and knowledge distillation, which secured 3rd place in CVPR OpenEarthMap Few-shot Challenge.
	Query Adaptive Transformer and Multiprototype Rectification for Few-Shot Remote Sensing Image Segmentation Tianyi Gao, Zhi Gao, Hong Ji, Wei Ao, Weiwei Song IEEE Transactions on Geoscience and Remote Sensing (TGRS), 2024 We propose a few-shot segmentation method that generates segmentor weights per-query using query-aware feature extraction and prototype rectification.
	Prompting-to-Distill semantic knowledge for few-shot learning Hong Ji, Zhi Gao, Jinchang Ren, Xing-ao Wang, Tianyi Gao, Wenbo Sun, Ping Ma IEEE Geoscience and Remote Sensing Letters (GRSL), 2024
	Adapting Vision Transformer for Few-Shot Remote Sensing Image Segmentation: Synergizing In-Domain Representations And Pretrained Model Guidance Tianyi Gao, Zhi Gao, IEEE International Geoscience and Remote Sensing Symposium (IGARSS, Oral), 2024
	Dual-modality vehicle anomaly detection via bilateral trajectory tracing Jingyuan Chen, Guanchen Ding, Yuchen Yang, Wenwei Han, Kangmin Xu, Tianyi Gao, Zhe Zhang, Wanping Ouyang, Hao Cai, Zhenzhong Chen IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (AI City Challenge), 2021

Services

Conference Reviewer: ICLR 2026 ML4RS, IGARSS 2026, WACV 2026 FoMoV, NeurIPS 2025, IGARSS 2025
Journal Reviewer: IEEE Transactions on Circuits and Systems for Video Technology (IF=11.1), Journal of Remote Sensing (IF=8.8), Remote Sensing Letters (IF=1.5)

Miscellaneous

I love the beauty of nature and delicious food 🍃🍜. I used to capture moments through photos, but now I just prefer to feel them with my heart (and also my stomach).
I enjoy playing basketball and fitness in my spare time. Reading and bouldering are also on my trying-to-develop list, with a rather modest revisit frequency 🛰️🌍
After engaging with many wonderful people during my time at MSRA, I eventually pulled myself back from the edge of job hunting and dived straight into the PhD grind 😁

This webpage is adapted from Jon Barron's page.