[ICLR2023] TaskPrompter on Cityscapes-3D: Experiment of Joint 2D-3D Multi-Task Prediction

Описание к видео [ICLR2023] TaskPrompter on Cityscapes-3D: Experiment of Joint 2D-3D Multi-Task Prediction

This is a demonstration video of our accepted ICLR 2023 research paper: TaskPrompter: Spatial-Channel Multi-Task Prompting for Dense Scene Understanding, authored by Hanrong Ye and Dan Xu from HKUST.

Trained on Cityscapes-3D, TaskPrompter is capable of simultaneously executing multiple key scene understanding tasks essential for the perception system of autonomous driving, including monocular 3D vehicle detection, semantic segmentation, and monocular depth estimation.

Project site: https://github.com/prismformore/Multi-Task...

Комментарии

Информация по комментариям в разработке

Похожие видео