A new Virtual Multi-View Synthesis module improves pedestrian orientation estimation when integrated into the AVOD-FPN 3D detector, outperforming prior methods on KITTI Orientation, 3D, and Bird's Eye View benchmarks.
Deep manta: A coarse-to-fine many-task network for joint 2d and 3d vehicle analysis from monocular image
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Improving 3D Object Detection for Pedestrians with Virtual Multi-View Synthesis Orientation Estimation
A new Virtual Multi-View Synthesis module improves pedestrian orientation estimation when integrated into the AVOD-FPN 3D detector, outperforming prior methods on KITTI Orientation, 3D, and Bird's Eye View benchmarks.