ToFNest: Efficient Normal Estimation for Time-of-Flight Depth Cameras

Szilárd Molnár, Benjamin Kelényi, Levente Tamás; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2021, pp. 1791-1798

In this work, we propose an efficient normal estimation method for depth images acquired by Time-of-Flight (ToF) cameras based on feature pyramid networks (FPN). We perform the normal estimation starting from the 2D depth images, projecting the measured data into the 3D space and computing the loss function for the point cloud normal. Despite the simplicity of our method, which we call ToFNest, it proves to be efficient in terms of robustness and runtime. In order to validate ToFNest we performed extensive evaluations using both public and custom outdoor datasets. Compared with the state of the art methods, our algorithm is faster by an order of magnitude without losing precision on public datasets. The demo code, custom datasets and videos are available on the project website.