First training stage of I2IwFiLM: a) Main U-Net model predicts ${\hat{\mathcal{I}}}_{\mathcal{C}}$ from input ${\hat{\mathcal{I}}}_{\mathcal{W}}$ and from ℰ^𝒫, the guidance vector extracted from the pairs 𝒫 := (ℐ_𝒲, ℐ_𝒞) using a guidance vector prediction network, denoted GVP_p. b) Details on the modulation of an intermediate feature map of the U-Net model by ℰ^𝒫 using a FiLM layer. The height, width, and number of channels (depth) of the intermediate feature map are denoted W, H, and C, respectively.