Ask HN: What are some unexplored or poorly explored applications of Comp Vision?

I'm asking this from two angles:

- What areas are current models and current approaches not sufficient?

- Is there any chance that the job market for these areas will exist if one chooses to specialize in Comp. Vision?