An iterative multi-task GAN-based framework completes occluded vehicle segmentation masks and recovers invisible appearance using coupled discriminators, a 3D silhouette pool, and a shared two-path network, outperforming prior methods on a new synthetic-plus-real dataset.
Title resolution pending
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
-
Visualizing the Invisible: Occluded Vehicle Segmentation and Recovery
An iterative multi-task GAN-based framework completes occluded vehicle segmentation masks and recovers invisible appearance using coupled discriminators, a 3D silhouette pool, and a shared two-path network, outperforming prior methods on a new synthetic-plus-real dataset.