I wonder if there would be any benefit for ML models to non linearly distorting image pixel statistics into a true Gaussian distribution instead of just the traditional scale and bias.

→ View original post on X — @id_aa_carmack