21
When it was created in 2017, the project emerged to support military analysts in the face of the avalanche of images generated by drones. Until then, the work was done manually: operators had to examine frame by frame to identify possible clues — often visible for just a moment — in a time-consuming and error-prone process.
