Wind current estimation from temperature by CNN

Arrows visualize the difference between the ground truth velocity field and velocity field from CNN. 3D tensor of temperatures is fed into the network, which is then trained by computing loss between reference trajectories (red color) and trajectories generated from CNN. The trajectories are generated by integrating simple motion model of a plane, with known control inputs and initial state, thus only velocity field remains unknown.

[@evamusilova7518]

love it bestie💅💅