w.r.t the material, it seems to be possible, since ToF sensors operate in a narrow 850 nm band. However, I don't know if a coating can be made that also doesn't

w.r.t the material, it seems to be possible, since ToF sensors operate in a narrow 850 nm band. However, I don't know if a coating can be made that also doesn't overlap with the visible light range since 850 nm is somewhat close to red light, or whether that would be cheap if so (given that the hidden cameras themselves are only $1 at the moment). I also have similar reservations about the DL approach in general. What makes me more confident is that we trained on a relatively small number of objects (~10) and tested on a totally different set of 30 objects. We're also using very small regions-of-interest as inputs (5x5), so there's little to no overfitting on the environment or objects themselves.