How do you select the best model?

I'm trying to reproduce the results reported in the paper. Thanks the authors for releasing the code! One question here, since all of the images in the training set are used for training (like in ShanghaiTechA), how do you select the best model? Any replies are appreciated, thanks!