In recent years, with the advent of deep-learning, face recognition has\nachieved exceptional success. However, many of these deep face recognition\nmodels perform much better in handling frontal faces compared to profile faces.\nThe major reason for poor performance in handling of profile faces is that it\nis inherently difficult to learn pose-invariant deep representations that are\nuseful for profile face recognition. In this paper, we hypothesize that the\nprofile face domain possesses a latent connection with the frontal face domain\nin a latent feature subspace. We look to exploit this latent connection by\nprojecting the profile faces and frontal faces into a common latent subspace\nand perform verification or retrieval in the latent domain. We leverage a\ncoupled conditional generative adversarial network (cpGAN) structure to find\nthe hidden relationship between the profile and frontal images in a latent\ncommon embedding subspace. Specifically, the cpGAN framework consists of two\nconditional GAN-based sub-networks, one dedicated to the frontal domain and the\nother dedicated to the profile domain. Each sub-network tends to find a\nprojection that maximizes the pair-wise correlation between the two feature\ndomains in a common embedding feature subspace. The efficacy of our approach\ncompared with the state-of-the-art is demonstrated using the CFP, CMU\nMulti-PIE, IJB-A, and IJB-C datasets. Additionally, we have also implemented a\ncoupled convolutional neural network (cpCNN) and an adversarial discriminative\ndomain adaptation network (ADDA) for profile to frontal face recognition. We\nhave evaluated the performance of cpCNN and ADDA and compared it with the\nproposed cpGAN. Finally, we have also evaluated our cpGAN for reconstruction of\nfrontal faces from input profile faces contained in the VGGFace2 dataset.\n