With the advent of deep learning architectures, the performance of face recognition has witnessed significant improvements. However, this has also necessitated the requirement of large labeled training database. While approaches exist to utilize labeled or unlabeled data from related domains, in this paper, we present a collaborative learning framework that utilizes the availability of both labeled and unlabeled data along with the presence of multiple experts, to improve the performance of face analysis related tasks. The proposed Label Consistent Deep Collaborative Learning (LC-DECAL) framework makes use of label consistency, transfer learning, ensemble learning, and co-training for training a deep neural network for the target domain. The efficacy of the proposed algorithm is demonstrated with two existing Convolutional Neural Network architectures, DenseNet and ResNet, via experiments on multiple face databases, namely YTF, PaSC Handheld, PaSC Control, CelebA, and LFW-a. Experimental results show that the proposed framework yields comparable results to state-of-the-art results on all the databases. © 2019 IEEE.