Abstract
In this paper, we present a real-time object detection and depth estimation approach based on deep convolutional neural networks (CNNs). We improve object detection through the incorporation of transfer connection blocks (TCBs), in particular, to detect small objects in real time. For depth estimation, we introduce binocular vision to the monocular-based disparity estimation network, and the epipolar constraint is used to improve prediction accuracy. Finally, we integrate the two-dimensional (2D) location of the detected object with the depth information to achieve real-time detection and depth estimation. The results demonstrate that the proposed approach achieves better results compared to conventional methods.
Highlights
Autonomous driving techniques [1,2,3] have been studied intensively for several decades
The stereo camera system used for data acquisition cost much less than light detection and ranging (LiDAR) or other time of flight (ToF) sensors
We presented an object detection and depth estimation approach based on deep learning techniques
Summary
Autonomous driving techniques [1,2,3] have been studied intensively for several decades. The Society of Automotive Engineers classifies five levels of automated driving, the third level of which is conditional automation, or self-driving under ideal conditions with limitations This level has drawn much attention as developers attempt to implement effective detection and recognition of the surrounding environment (e.g., the road, traffic signs, other vehicles, and pedestrians) so that the vehicle can detect and recognize objects ahead and estimate their depth from a visual sensor. We propose a real-time object detection and depth estimation approach using learning-based techniques for images acquired from a vehicle’s onboard camera. We present an improved object detection approach—in particular for small objects—and use deep neural networks and epipolar geometry to create stereo images and generate depth maps.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
