GS3D: An Efficient 3D Object Detection Framework for Autonomous Driving

Abstract

We present an efficient 3D object detection framework based on a single RGB image in the scenario of autonomous driving. Our efforts are put on extracting the underlying 3D information in a 2D image and determining the accurate 3D bounding box of the object without point cloud or stereo data. Leveraging the off-the-shelf 2D object detector, we propose an artful approach to efficiently obtain a coarse cuboid for each predicted 2D box. The coarse cuboid has enough accuracy to guide us to determine the 3D box of the object by refinement. In contrast to previous state-ofthe-art methods that only use the features extracted from the 2D bounding box for box refinement, we explore the 3D structure information of the object by employing the visual features of visible surfaces. The new features from surfaces are utilized to eliminate the problem of representation ambiguity brought by only using a 2D bounding box. Moreover, we investigate different methods of 3D box refinement and discover that a classification formulation with quality aware loss has much better performance than regression. Evaluated on the KITTI benchmark, our approach outperforms current state-of-the-art methods for single RGB image based 3D object detection.

PointRCNN: 3D Object Proposal Generation and Detection From Point Cloud

“…There are existing works on estimating the 3D bounding box from images. [24,15] leveraged the geometry constraints between 3D and 2D bounding box to recover the 3D object pose. [1,44,23] exploited the similarity between 3D objects and the CAD models.…”

A review on object pose recovery: From 3D bounding box detectors to full 6D pose estimators

“…6D object pose estimators [27], [28], [29], [33] extract features from the input images (feature extraction block), and using the trained classifiers, estimate objects' 6D pose. Several methods further refine the output of the trained classifiers [104], [81], [149], [28], [29], [33] (refinement block), and finally hypothesise the object pose after filtering. Table II details the classification-based methods.…”

Fast and Reliable Alternative to Encoder-Based Measurements of Multiple 2-DOF Rotary-Linear Transformable Objects Using a Network of Image Sensors with Application to Table Football

“…This results in improved performance over constantly running object detection, but requires a reliable failure detection and recovery [ 21 , 22 ]. The reliability of the detector and tracker is of paramount importance for automotive applications [ 23 ], where incorrect object position or the orientation can result in dangerous reaction of the automated driving system. Other proposed solutions use object model for robust tracking in complex environments [ 24 ], the idea that is used and enhanced in our approach.…”