Few-shot Semantic Learning for Robust Multi-Biome 3D Semantic Mapping in Off-Road Environments
Atha, Deegan, Lei, Xianmei, Khattak, Shehryar, Sabel, Anna, Miller, Elle, Noca, Aurelio, Lim, Grace, Edlund, Jeffrey, Padgett, Curtis, Spieler, Patrick
–arXiv.org Artificial Intelligence
Off-road environments pose significant perception challenges for high-speed autonomous navigation due to unstructured terrain, degraded sensing conditions, and domain-shifts among biomes. Learning semantic information across these conditions and biomes can be challenging when a large amount of ground truth data is required. In this work, we propose an approach that leverages a pre-trained Vision Transformer (ViT) with fine-tuning on a small (<500 images), sparse and coarsely labeled (<30% pixels) multi-biome dataset to predict 2D semantic segmentation classes. These classes are fused over time via a novel range-based metric and aggregated into a 3D semantic voxel map. We demonstrate zero-shot out-of-biome 2D semantic segmentation on the Yamaha (52.9 mIoU) and Rellis (55.5 mIoU) datasets along with few-shot coarse sparse labeling with existing data for improved segmentation performance on Yamaha (66.6 mIoU) and Rellis (67.2 mIoU). We further illustrate the feasibility of using a voxel map with a range-based semantic fusion approach to handle common off-road hazards like pop-up hazards, overhangs, and water features.
arXiv.org Artificial Intelligence
Nov-10-2024
- Country:
- North America > United States (1.00)
- Genre:
- Research Report (0.64)
- Industry:
- Technology:
- Information Technology > Artificial Intelligence
- Machine Learning (1.00)
- Natural Language > Text Processing (0.35)
- Representation & Reasoning (1.00)
- Robots (0.96)
- Vision (0.88)
- Information Technology > Artificial Intelligence