Yinda Zhang
Papers
4
Total Citations
295
H-Index
4
About
Yinda Zhang is a leading researcher in computer vision and graphics, whose work bridges the gap between physically-based simulation and deep learning for scene understanding. His most influential contribution, "Physically-Based Rendering for Indoor Scene Understanding Using Convolutional Neural Networks" (2017, 277 citations), pioneered the use of synthetic data generated by physically-accurate rendering to train CNNs for indoor scene analysis—a critical enabler for robot navigation and human-assistive AI. This approach addresses the bottleneck of costly real-world annotation by leveraging photorealistic synthetic environments, setting a new standard for data-efficient training in vision. More recently, Zhang has advanced dynamic scene understanding with "GaussianPrediction: Dynamic 3D Gaussian Prediction for Motion Extrapolation and Free View Synthesis" (2024), which forecasts future scenarios in dynamic environments—a key capability for intelligent navigation and decision-making in robotics. His earlier work on joint hand detection and rotation estimation (2016) also demonstrates his versatility in tackling human-computer interaction challenges. With over 290 total citations and a growing impact, Zhang’s research continues to shape how machines perceive and interact with complex, dynamic worlds, making him a notable figure in modern computer vision.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4Joint Hand Detection and Rotation Estimation by Using CNN4 citations · 2016