Hsiao-Yu Fish Tung

Papers

Total Citations

H-Index

About

Hsiao-Yu Fish Tung is a leading researcher in computer vision and robotics, whose work bridges the gap between 3D scene understanding and intelligent manipulation. Her core research areas include 3D neural scene representations, object-centric learning, and fluid manipulation. Tung’s major contribution is the development of **3D-OES** (2020, 13 citations), a pioneering viewpoint-invariant dynamics model that predicts scene changes from RGB-D video by factorizing objects in a 3D neural space—enabling robots to reason about object interactions without visual interference. She further advanced robotic dexterity with **FluidLab** (2023, 11 citations), a differentiable simulation environment that benchmarks complex fluid tasks like scooping and pouring, tackling the long-standing challenge of non-rigid object manipulation. Her work on **Reward Learning from Narrated Demonstrations** (2018, 6 citations) introduced a novel method for robots to infer goals from natural language, moving beyond traditional visual or pose-based programming. Tung’s research has been recognized for its impact on embodied AI, earning her citations across top venues. Her contributions are shaping how robots perceive, predict, and interact with dynamic, real-world environments—from solid objects to complex fluids.