- Collaborative Semantic Occupancy Prediction with Hybrid Operate Fusion in Linked Automated Autos(arXiv)
Author : Rui Song, Chenwei Liang, Hu Cao, Zhiran Yan, Walter Zimmer, Markus Gross, Andreas Festag, Alois Knoll
Abstract : Collaborative notion in automated cars leverages the alternate of data between brokers, aiming to boost notion outcomes. Earlier camera-based collaborative 3D notion methods often make use of 3D bounding containers or chook’s eye views as representations of the setting. Nonetheless, these approaches fall fast in offering an entire 3D environmental prediction. To bridge this gap, we introduce the first approach for collaborative 3D semantic occupancy prediction. Considerably, it improves native 3D semantic occupancy predictions by hybrid fusion of (i) semantic and occupancy exercise choices, and (ii) compressed orthogonal consideration choices shared between cars. Furthermore, due to the shortage of a collaborative notion dataset designed for semantic occupancy prediction, we improve a gift collaborative notion dataset to include 3D collaborative semantic occupancy labels for a additional sturdy evaluation. The experimental findings highlight that: (i) our collaborative semantic occupancy predictions excel above the outcomes from single cars by over 30%, and (ii) fashions anchored on semantic occupancy outpace state-of-the-art collaborative 3D detection methods in subsequent notion features, showcasing enhanced accuracy and enriched semantic-awareness in road environments.
