Table 1 Class definitions and cutoff points that we developed and used to manually label the StairNet data set [13]
From: StairNet: visual recognition of stairs for human–robot locomotion
StairNet class | ExoNet class | Class example | Class description |
|---|---|---|---|
LG | LG Steady State, LG-Door/Wall |
| An image that contains a level ground environment where incline stairs are not clearly visible |
LG–IS | LG–IS |
| An image with incline stairs where the horizontal surface area of the bottom step or landing is clearly greater than the surface area of other steps visible in the image (i.e., the surface area or depth is approximately 1.5 × the size of subsequent steps) |
IS | IS Steady State, IS-Door/Wall |
| An image with multiple incline stairs where the horizontal surface area of the top and bottom step or landing is not clearly greater than one another |
IS–LG | IS–LG |
| An image with incline stairs where the horizontal surface area of the top step or landing is clearly greater than that of other steps or landings visible in the image (i.e., the surface area or depth is approximately 1.5 × the size of subsequent steps). For an incline stair to be included in the IS–LG class, the horizontal face of the last step prior to level ground must be visible |
