Multimodal Terrain Classification for Off-Road Environments

Jul 15, 2021
RGBD Dataset Sample

  • Built a robust terrain classification model using depth assisted semantic segmentation.
  • Collected and generated a novel RGBD + Semantic Segmentation dataset of completely off-road / wild terrains.
  • Employed the use of SOTA models like Intel’s Dense Prediction Transformer and MiDaS for dataset benchmarking.
  • Dual-stream RGB + Depth channel based multi-modal architecture with fusion-based feature combination.