Indoor Navigation 2D Data Set

In2d all envs.png

The dataset consists of 17 2D maps of indoor office and living environments generated from a subset of the Gibson Database, which contains 3D scans of real environments. The maps also have .yaml files for direct loading with the ROS package map_server. The dataset additionally contains start and goal coordinates for evaluating navigation tasks. Features of the dataset:

  • 17 environments represented as images with a resolution of 1cm per pixel.
  • 1700 short navigation tasks (100 per environment), with the distance between start and goal (0,4]m and at least one obstacle in between.
  • 425 long navigation tasks (25 per environment), with random distances between start and goal, with intermediate goals set at a distance of 1m.


If you use this dataset please cite our paper:

  title={Adaptive Dynamic Window Approach},
  author={Dobrevski, Matej and Sko{\v{c}}aj, Danijel},
  booktitle={2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},

Additionally please cite the Gibson Database, from which our dataset was generated:

  title={Gibson {Env}: real-world perception for embodied agents},
  author={Xia, Fei and R. Zamir, Amir and He, Zhiyang and Sax, Alexander and Malik, Jitendra and Savarese, Silvio},
  booktitle={Computer Vision and Pattern Recognition (CVPR), 2018 IEEE Conference on},


The dataset can be downloaded here.