This document summarizes research on developing an algorithm for thermal-to-visible light face recognition. The algorithm performs preprocessing of images, uses a convolutional neural network to extract features from thermal and visible light images of the same individuals, and compares the images using a cost function to identify matches across modalities. The algorithm is still being developed and evaluated on its ability to accurately match faces between visible light and thermal images. The goal is to refine the cost function and neural network to improve the identification rate for cross-modal face recognition.