Dataset for training a semantic image segmentation model for the Intelligent Ground Vehicle Competition.
Each instance consists of an reference image from the point of view of the robot and the corresponding obstacle (e.g. construction drums, buckets) and lane segmentation masks.
Train
256 frames rendered in 4 different lighting environments using Blender = 1024 images
Test
10 frames captured from the SCR 2023 IGVC run (manually… See the full description on the dataset page:
https://huggingface.co/datasets/ngres/IGVC-Segmentation.
