Figure 6.

Image post-processing steps. (a) Source image from the color camera (left) and its semantic segmentation (right); (b) pixels not labeled as persons are filtered out; (c) pixels beyond 1.5 m are filtered out; (d) pixels closer than 5 m are filtered out. This includes pixels whose depth value cannot be calculated (reported distance 0 m); (e) Of the remaining pixels, only the $25\%$ closest to the camera are retained, which are the pixels that contribute to the sonification.