Figure 3
From: Pseudo contrastive labeling for predicting IVF embryo developmental potential
Models overview. (A) A CNN model, trained in a fully supervised manner over KIDp versus KIDn, a total of 7799 embryos. (B) A temporal ARMA filtering over the outcome of A. (C) A CNN model, trained using pseudo constrictive labels obtained by the outcome of B, over all training embryos, a total of 61581 embryos (D) An ARMA filtering over the outcome of C. (E) A second iteration of pseudo constrictive labels: a CNN model, trained using pseudo constrictive labels obtained by the outcome of D. (F) An ARMA filtering over the outcome of E.
