Semantic Scene Completion in Autonomous Driving: A Two-Stream Multi-Vehicle Collaboration Approach
Junxuan Li, Yuanfang Zhang, Jiayi Han, Peng Han, Kaiqing Luo, Yilong Hui, Wei Shao · 2024
Read Semantic Scene Completion in Autonomous Driving: A Two-Stream Multi-Vehicle Collaboration Approach by Junxuan Li, Yuanfang Zhang, Jiayi Han, Peng Han, Kaiqing Luo, Yilong Hui, Wei Shao (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-11644936.txt.
Description
Junxuan Li and six colleagues let self-driving cars pool their views to complete occluded street scenes, in a paper from two thousand twenty-four. A single vehicle cannot see past trucks and around corners. Connected vehicles can share sensor data. But to align the features in that data is hard.
Their two-stream method splits partner vehicles' features apart and lets them interact. A custom attention module lets the host car query the most similar local features from its partners. It does so in place of assuming perfect synchronization of the feeds. The fused features give rise to a three-dimensional occupancy map. That map carries semantic labels for what sits in each place. Two multi-vehicle datasets benchmarked the approach for the team. Cooperative perception holds out the hope of safer autonomy. That holds if alignment errors do not invent phantom obstacles, or wipe real ones out.
Catalog record
| CREATOR | Junxuan Li, Yuanfang Zhang, Jiayi Han, Peng Han, Kaiqing Luo, Yilong Hui, Wei Shao |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |