Confidence score distributions for DBNets2.0 estimates on the synthetic observations of the test set randomly augmented with the same augmentation layers used during training: random rotation, random translation, random cut of the outer disc, convolution with a Gaussian beam of random size and addition of white noise. The blue histogram corresponds to in-distribution (ID) augmentation, i.e. using the same hyperparameters used during training. The orange histogram corresponds to values obtained for images on which we performed OOD augmentation increasing the maximum translation factor from 0.01 to 0.1, the noise standard deviation from 0.1 to 0.2 and the maximum synthetic beam size from 0.2 to 0.4 (in code units).