Deep Learning based 3D Object Detection in Indoor Environments: A Review

摘要

Recently, the performance of object detection models have been efficiently improved with the application of deep learning in point clouds. However, as far as we know, most proposed reviews focus on outdoor scenes for autonomous driving. So in this paper, we provide a comprehensive review of 3D object detection for point clouds in cluttered indoor environments, which is widely used in the fields of robotics and augmented reality. Firstly, we introduce three most frequently used indoor datasets. Then, we review the representative detection models in recent years and sort these methods into two classifications, segmentation-based models and non-segmentation models. The characteristics of each method are summarized and the results are compared on three different datasets. Lastly, we conclude the insightful observations and future works.