Non-parametric morphology methods, such as CAS, are powerful fast measurements to estimate the structure and morphology of galaxies. These indicators are particularly useful when selecting sub samples of galaxy mergers from a larger population, enabling us to quickly measure merger fractions and merger rates using only the morphology. However, there are distinct types of morphologies that have the potential to result in similar non-parametric morphology measurements. This is the case of Post-Mergers (PM) and Non-Merging Star Forming galaxies (NM-SF), where this ambiguity results in contamination in merger samples selected only through morphology, introducing a source of uncertainty in merger fractions. Here we show that it is possible to use deep learning models to distinguish these two types of galaxies. To do so, we generated a balanced dataset of images of post mergers and non-merging star forming galaxies from the IllustrisTNG simulation using a full radiative transfer treatment, including dust. Then, using the merger trees available from the simulation, we label the images as PM or NM-SF. We are able to train a deep learning model that reaches 90% accuracy when distinguishing these two types of galaxies within IllustrisTNG.