PlaneSegNet: Fast and Robust Plane Estimation Using a Single-stage Instance Segmentation CNN
This work addresses the need for fast and robust plane estimation for applications like visual SLAM and augmented reality, representing an incremental improvement over prior methods.
The authors tackled the problem of slow frame rates in existing two-stage methods for instance segmentation of planar regions in indoor scenes by proposing PlaneSegNet, a real-time deep neural architecture that achieves significantly higher frame-rates with comparable segmentation accuracy.
Instance segmentation of planar regions in indoor scenes benefits visual SLAM and other applications such as augmented reality (AR) where scene understanding is required. Existing methods built upon two-stage frameworks show satisfactory accuracy but are limited by low frame rates. In this work, we propose a real-time deep neural architecture that estimates piece-wise planar regions from a single RGB image. Our model employs a variant of a fast single-stage CNN architecture to segment plane instances. Considering the particularity of the target detected, we propose Fast Feature Non-maximum Suppression (FF-NMS) to reduce the suppression errors resulted from overlapping bounding boxes of planes. We also utilize a Residual Feature Augmentation module in the Feature Pyramid Network (FPN). Our method achieves significantly higher frame-rates and comparable segmentation accuracy against two-stage methods. We automatically label over 70,000 images as ground truth from the Stanford 2D-3D-Semantics dataset. Moreover, we incorporate our method with a state-of-the-art planar SLAM and validate its benefits.