The pick and place task, is one of the most frequently performed tasks by robots in the industrial world. In typical production plants, the objects locations are known in advance and do not change, i.e. the handled products are placed in specialized holders, which prevent them from changing their position and orientation even if a disturbing force occurs. This method has the advantage of ensuring robustness, but when scaled up, the holders add additional complexity and increase the cost of installation and implementation. To satisfy an increasingly diverse demand and product range, manufacturers are requiring more flexibility in their installations. Computer vision-based localization and pose estimation solutions are therefore being proposed by technology operators to overcome the limitations of fixed holders and make pick and place applications more flexible. A multi-robot cell would then not only allow to reach objectives that are unattainable for a single robot, but would also increase the overall performance thanks to the parallelization of complementary skills of the robots.