Saliency map of ASID-FE’s uncertainties prediction for a single source. The images are from left to right: original image, saliency for σ_x, saliency for σ_y and saliency for σ_flux. The colour map indicates the degree of importance assigned to each pixel in the image. The brighter the colour, the more important the pixel is for the prediction. Results suggest that the network focuses mostly on the central pixels to estimate σ_x and σ_y, while it pays more attention to a wide area around the source to estimate σ_flux.