Occlusion-Aware Road Layout Prediction from Perspective RGB Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for understanding road layout in top view from perspective input require expensive pixel-level annotations and fail to reason occlusion relationships effectively, limiting their applicability in autonomous driving and path planning.
Innovation Solution
A computer-implemented method that segments RGB images into pixel-level semantic segmentation results in perspective view, learns a mapping to top view using a road plane assumption, and generates occlusion-aware parametric road layout predictions, reducing the need for expensive annotations through deep supervision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pixel-level semantic segmentation is performed in perspective view to understand road layout, then measurement precision of road layout is improved, but loss of time and resources for annotation increases
Solution Approach 1:
The annotation process is segmented into two levels: (1) coarse parametric annotations for road layout attributes (lane positions, widths, orientations) that are efficient to annotate, and (2) automatic generation of detailed pixel-level semantic segmentation from these parametric annotations. This segmentation allows obtaining precise pixel-level results without manually annotating every pixel, thus reducing annotation time while maintaining measurement precision.
Solution Approach 2:
The system performs preliminary action by first obtaining coarse parametric road layout annotations, then automatically generating the detailed pixel-level semantic segmentation maps as preprocessing steps. This preliminary generation of pixel-level labels from parametric annotations eliminates the need for time-consuming manual pixel-level annotation while ensuring high measurement precision for road layout understanding.
2Measurement precision
If pixel-level semantic segmentation is used for road layout understanding, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system segments the processing into distinct modules: (1) a module for obtaining parametric road layout annotations, (2) a generation module that converts parametric annotations to pixel-level semantic segmentation, and (3) a module for reasoning occlusion relationships. This modular segmentation manages device complexity by organizing functions into separate, manageable components rather than a monolithic complex system.
Solution Approach 2:
The system introduces an intermediary representation (parametric road layout attributes) that bridges the gap between simple input data and complex pixel-level semantic segmentation outputs. This intermediary layer simplifies the overall system complexity by providing a structured intermediate form that is easier to process and reason about compared to direct pixel-level processing.
3Reliability
If top-view representations are generated to reason occlusion relationships, then reliability of occlusion reasoning is improved, but manufacturing precision of mapping increases
Solution Approach 1:
The system changes parameters by representing road layout in two different coordinate systems: perspective view parameters for visible regions and top-view parameters for occluded regions. This parameter transformation allows reliable occlusion reasoning in top-view while managing the complexity of perspective-to-top-view mapping through established geometric transformations and road plane assumptions.
Data Source
AI summary
A computer-implemented method for road layout prediction is provided. The method includes segmenting, by a first processor-based element, an RGB image to output pixel-level semantic segmentation results for the RGB image in a perspective view for both visible and occluded pixels in the perspective view based on contextual clues. The method further includes learning, by a second processor-based element, a mapping from the pixel-level semantic segmentation results for the RGB image in the perspective view to a top view of the RGB image using a road plane assumption. The method also includes generating, by a third processor-based element, an occlusion-aware parametric road layout prediction for road layout related attributes in the top view.