Scene-aware Data Augmentation
Ken Yu's scene-aware data augmentation research improves training data for computer vision and autonomous driving models.
Summary
This research project studies scene-aware data augmentation for computer vision. Ken Yu explored how to place copied instances into new scenes more realistically so perception models can learn from richer and more plausible synthetic training examples.
Focus areas: data augmentation, scene understanding, instance placement, perception model training.
Motivation
Related Work
- Panoptic-DepthLab for segmentation and depth prediction.
- Perspective-aware Convolution for monocular 3D object detection.
- Publications for related research and thesis output.
Introduction
Approach
Result
