When we presented our work at the Vision 2007, we were happy to meet many very experienced people, who thought that we are exaggerating a little bit – they thought that there is no need (other than academic) to fully reconstruct images one by one, independently from each other.
We followed these suggestions and created a program version, which:
- subtracts the constant background scene
- reconstructs humans in more detail but is still faster than the previous version
- has improved distance measurement
- gives an alarm, if the human puts up a weapon