IAFA: Instance-aware Feature Aggregation for 3D Object Detection from a Single Image

03/05/2021
by   Dingfu Zhou, et al.
0

3D object detection from a single image is an important task in Autonomous Driving (AD), where various approaches have been proposed. However, the task is intrinsically ambiguous and challenging as single image depth estimation is already an ill-posed problem. In this paper, we propose an instance-aware approach to aggregate useful information for improving the accuracy of 3D object detection with the following contributions. First, an instance-aware feature aggregation (IAFA) module is proposed to collect local and global features for 3D bounding boxes regression. Second, we empirically find that the spatial attention module can be well learned by taking coarse-level instance annotations as a supervision signal. The proposed module has significantly boosted the performance of the baseline method on both 3D detection and 2D bird-eye's view of vehicle detection among all three categories. Third, our proposed method outperforms all single image-based approaches (even these methods trained with depth as auxiliary inputs) and achieves state-of-the-art 3D detection performance on the KITTI benchmark.

READ FULL TEXT

page 2

page 14

research
08/22/2021

SIDE: Center-based Stereo 3D Detector with Structure-aware Instance Depth Estimation

3D detection plays an indispensable role in environment perception. Due ...
research
08/14/2014

2D View Aggregation for Lymph Node Detection Using a Shallow Hierarchy of Linear Classifiers

Enlarged lymph nodes (LNs) can provide important information for cancer ...
research
11/26/2018

MonoGRNet: A Geometric Reasoning Network for Monocular 3D Object Localization

Localizing objects in the real 3D space, which plays a crucial role in s...
research
03/09/2020

BirdNet+: End-to-End 3D Object Detection in LiDAR Bird's Eye View

On-board 3D object detection in autonomous vehicles often relies on geom...
research
12/11/2022

Focal-PETR: Embracing Foreground for Efficient Multi-Camera 3D Object Detection

The dominant multi-camera 3D detection paradigm is based on explicit 3D ...
research
01/19/2015

Coupled Depth Learning

In this paper we propose a method for estimating depth from a single ima...
research
09/17/2020

Single Frame Deblurring with Laplacian Filters

Blind single image deblurring has been a challenge over many decades due...

Please sign up or login with your details

Forgot password? Click here to reset