Examples of cutouts used to train the neural network Top: examples of simulated lenses based on real DES images. Stamps are ordered by increasing Einstein radii (top-left to bottom-right). The top row corresponds to θ_E = 1.2–1.8″, the middle row to θ_E = 1.8–2.4″, and the bottom row to θ_E = 2.4–3.0″. Bottom: examples of LRGs used either as non-lenses during the training of the CNNs (see Sect. 4.1) or as objects onto which we inject a lensed source to build simulated lens systems, as shown in the top panel. All the cutouts are 50 pixels on-a-side, corresponding to 13′′..