Authors = {Nazrul Haque, Dinesh Reddy and K. Madhava Krishna} Robotics Research Center, IIIT Hyderabad, India. - http://robotics.iiit.ac.in/
The repository contains semantic Ground Truth annotations for 200 images from KITTI Tracking dataset[Geigar].
The zip file contains two directories.
annotated-dataset
|--RGB - Contains RGB images from KITTI Tracking dataset (Geigar)
|--GT - Contains pixel-wise Ground Truth annotations for 11 semantic classes.
The Ground Truth label spectrum is given below:
| Class | Colour | R | G | B |
|---|---|---|---|---|
| Building | Red | 128 | 0 | 0 |
| Vegetation | Yellow | 128 | 128 | 0 |
| Sky | Grey | 128 | 128 | 128 |
| Car | Purple | 64 | 0 | 128 |
| Sign | Salmon | 192 | 128 | 128 |
| Road | Pink | 128 | 64 | 128 |
| Pedestrian | Yellow-Brown | 64 | 64 | 0 |
| Fence | Grey-Purple | 64 | 64 | 128 |
| Pole | Light-Yellow | 192 | 192 | 128 |
| Sidewalk | Blue | 0 | 0 | 192 |
| Cyclist | Light-Blue | 0 | 128 | 192 |
References:
[1] A. Geiger, P. Lenz, and R. Urtasun, “Are we ready for autonomous driving? the KITTI vision benchmark suite”, CVPR, 2012 Link - http://www.cvlibs.net/datasets/kitti/
[2] G. Ros and S. Ramos and M. Granados and A. Bakhtiary and D. Vazquez and A.M. Lopez, "Vision-based Offline-Online Perception Paradigm for Autonomous Driving". Link - http://adas.cvc.uab.es/s2uad/