MachMap: End-to-End Vectorized Solution for Compact HD-Map Construction
This work addresses the problem of efficient and accurate high-definition map construction for autonomous driving systems, representing a strong specific gain rather than a foundational advancement.
The paper tackles HD-map construction for autonomous driving by introducing MachMap, an end-to-end vectorized solution that reduces vectorized points by 93% without performance loss and achieves a mAP of 83.5 on the Argoverse2 benchmark, outperforming other methods by at least 9.8 mAP.
This report introduces the 1st place winning solution for the Autonomous Driving Challenge 2023 - Online HD-map Construction. By delving into the vectorization pipeline, we elaborate an effective architecture, termed as MachMap, which formulates the task of HD-map construction as the point detection paradigm in the bird-eye-view space with an end-to-end manner. Firstly, we introduce a novel map-compaction scheme into our framework, leading to reducing the number of vectorized points by 93% without any expression performance degradation. Build upon the above process, we then follow the general query-based paradigm and propose a strong baseline with integrating a powerful CNN-based backbone like InternImage, a temporal-based instance decoder and a well-designed point-mask coupling head. Additionally, an extra optional ensemble stage is utilized to refine model predictions for better performance. Our MachMap-tiny with IN-1K initialization achieves a mAP of 79.1 on the Argoverse2 benchmark and the further improved MachMap-huge reaches the best mAP of 83.5, outperforming all the other online HD-map construction approaches on the final leaderboard with a distinct performance margin (> 9.8 mAP at least).