Tackling 3D ToF Artifacts Through Learning and the FLAT Dataset
This addresses depth reconstruction accuracy for users of time-of-flight cameras, such as in robotics or AR/VR, but is incremental as it builds on existing deep-learning methods.
The paper tackled artifacts in 3D time-of-flight depth reconstruction caused by scene motion, multiple reflections, and sensor noise, proposing a two-stage deep-learning approach and introducing the FLAT dataset, and showed improved reconstruction errors over state-of-the-art methods on simulated and real data.
Scene motion, multiple reflections, and sensor noise introduce artifacts in the depth reconstruction performed by time-of-flight cameras. We propose a two-stage, deep-learning approach to address all of these sources of artifacts simultaneously. We also introduce FLAT, a synthetic dataset of 2000 ToF measurements that capture all of these nonidealities, and allows to simulate different camera hardware. Using the Kinect 2 camera as a baseline, we show improved reconstruction errors over state-of-the-art methods, on both simulated and real data.