The data captured some short navigations by a person along some potential surroundings within the urban. These data are image-pair sequences captured with a rate of 30 frame per second, and stored as bitmap file format to preserve their best resolution. The image size is maintained at 752 x 480 pixels – the highest window allowed by our stereo camera.
The data are grouped into 4 taxonomy of (1) uncovered/partially covered drainage, (2) drop-off/uneven steps, (3) mix gradient, and (4) smooth surface. We collected the data with our prototype wearable device as mentioned in our Documentation.
We also measured the ground truth of the surface discontinuities in their 3D dimensions. For example, the following figure is the dimensions of some steps with a drop-off next to an uncovered drainage.

The following image-pairs are some examples of the said taxonomy:
uncovered / partially covered drainage –> request more samples
drop-off / uneven steps –> request more samples
mix gradient (steps + uncovered drainage) –> request more samples
smooth surface –> request more samples