Deep MANTA: A Coarse-to-fine Many-Task Network for joint 2D and 3D vehicle analysis from monocular image

被引：280

作者：

Chabot, Florian ^{[1
]}

Chaouch, Mohamed ^{[1
]}

Rabarisoa, Jaonary ^{[1
]}

Teuliere, Celine ^{[2
]}

Chateau, Thierry ^{[2
]}

机构：

[1] CEA LIST Vis & Content Engn Lab, Palaiseau, France

[2] Blaise Pascal Univ, Pascal Inst, Clermont Ferrand, France

来源：

30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017) | 2017年

关键词：

D O I：

10.1109/CVPR.2017.198

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

In this paper, we present a novel approach, called Deep MANTA (Deep Many-Tasks), for many-task vehicle analysis from a given image. A robust convolutional network is introduced for simultaneous vehicle detection, part localization, visibility characterization and 3D dimension estimation. Its architecture is based on a new coarse-to-fine object proposal that boosts the vehicle detection. Moreover, the Deep MANTA network is able to localize vehicle parts even if these parts are not visible. In the inference, the network's outputs are used by a real time robust pose estimation algorithm for fine orientation estimation and 3D vehicle localization. We show in experiments that our method outperforms monocular state-of-the-art approaches on vehicle detection, orientation and 3D location tasks on the very challenging KITTI benchmark.

引用

页码：1827 / 1836

页数：10