MonoSAOD: Monocular 3D Object Detection with Sparsely Annotated Label

Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• Monocular 3D object detection struggles when only a fraction of objects are labeled due to the high cost of 3D annotation.limitation 90%
• The proposed Road-Aware Patch Augmentation (RAPA) leverages sparse annotations by augmenting segmented object patches onto road regions while preserving 3D geometric consistency.method 95%
• The proposed Prototype-Based Filtering (PBF) generates high-quality pseudo-labels by filtering predictions through prototype similarity and depth uncertainty.method 95%
• Prototype-Based Filtering maintains global 2D RoI feature prototypes and selects pseudo-labels that are both feature-consistent with learned prototypes and have reliable depth estimates.technical 90%
• The training strategy combines geometry-preserving augmentation with prototype-guided pseudo-labeling to achieve robust detection under sparse supervision.method 90%
• The sparsely annotated setting is common in real-world scenarios where annotating every object is impractical.market 85%
• Extensive experiments demonstrate the effectiveness of the proposed method.result 80%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card