The facial paralysis is a medical condition that may difficult patients’ activities of daily living such as talk, eat, drink and close eyes. Further, such physical issues may carry future dental and/or ocular complications; for that reason, facial paralysis requires timely and precise treatment. The Deep Learning (DL), a field of Artificial Intelligence, has successfully incurred in the medical field supporting image-based diagnosis systems. Several models have been proposed to solve diseases like pneumonia, cancer, etc. Despite their success, obtaining an adequate model to successfully support the diagnosis of a specific disease may require large amounts of images, be time-consuming, or need specialized hardware. To take advance of capabilities of DL models, there exist techniques such as Transfer-Learning (TL) and Fine-Tuning (FT) that use pre-trained models on related or similar task for solving the interest task; thus, decreasing the requirements previously mentioned. In this work, there are used and compared TL and FT techniques applied to support the diagnosis of facial paralysis based on images.