We submit this supplementary material for the manuscript titled "Selective compression learning of latent representations for variable-rate image compression". This supplementary material includes:

(i) The extended manuscript including the appendix.
(ii) The CSV files that include the experimental results.
(iii) The video animations that show how the bpp value, PSNR value, reconstruction quality, generated 3D binary mask, and element selection ratio change as the target quality level q value changes. These animations are generated using the MSE-optimized SCR extension of the Hyperprior model. Note that we submit these animations in a compressed video format due to the size limit of the supplementary material.