无需相机参数即可预测图像平面3D框角点与深度，提升实际应用能力。
MoCA3D: Monocular 3D Bounding Box Prediction in the Image Plane
- 通过角点热图与深度图实现像素空间的密集预测。
- 图像平面角点一致性提升22.8%，参数量仅为基准模型1/57。
- 适用于相机参数未知场景，推动真实世界检测应用落地。
单目3D目标理解传统上被视为从2D区域到3D框的提升问题。然而，新兴下游应用需要图像平面几何信息（如投影3D框角点），在缺乏已知内参的情况下难以获取，限制了野外物体检测的应用。本文提出MoCA3D，一种单目、类别无关的3D模型，可在推理时无需相机内参，直接预测投影3D边界框角点及每个角点的深度。该方法将像素空间定位与深度分配建模为密集预测任务，通过角点热图和深度图实现。为评估图像平面几何保真度，提出像素对齐几何（PAG）指标，直接衡量角点与深度的一致性。大量实验表明，MoCA3D在图像平面角点PAG上提升22.8%，同时在3D IoU上保持相当性能，且参数量最多减少57倍。最后，将MoCA3D应用于此前因内参未知而不可行的下游任务，验证其超越标准基线模型的实际价值。
原文摘要 · Abstract (English)
Monocular 3D object understanding has largely been cast as a 2D RoI-to-3D box lifting problem. However, emerging downstream applications require image-plane geometry (e.g., projected 3D box corners) which cannot be easily obtained without known intrinsics, a problem for object detection in the wild. We introduce MoCA3D, a Monocular, Class-Agnostic 3D model that predicts projected 3D bounding box corners and per-corner depths without requiring camera intrinsics at inference time. MoCA3D formulates pixel-space localization and depth assignment as dense prediction via corner heatmaps and depth maps. To evaluate image-plane geometric fidelity, we propose Pixel-Aligned Geometry (PAG), which directly measures image-plane corner and depth consistency. Extensive experiments demonstrate that MoCA3D achieves state-of-the-art performance, improving image-plane corner PAG by 22.8% while remaining comparable on 3D IoU, using up to 57 times fewer trainable parameters. Finally, we apply MoCA3D to downstream tasks which were previously impractical under unknown intrinsics, highlighting its utility beyond standard baseline models.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。