Indoor localization has many pervasive applications, such as energy management, health monitoring, and security. Tagless localization detects directly the human body, for example via infrared sensing, and is the most amenable to different users and use cases. We evaluate the localization and tracking performance, as well as resource and processing requirements, of various neural network (NN) types. We use directly the data from a low-resolution 16-pixel thermopile sensor array in a 3 × 3 m room, without preprocessing or filtering. We tested several NN architectures, including multilayer perceptron, autoregressive, 1-D convolutional NN (1D-CNN), and long short-term memory. The latter require more resources but can accurately locate and capture best the person movement dynamics, whereas the 1D-CNN is the best compromise between localization accuracy (9.6-cm root-mean-square error), movement tracking smoothness, and required resources. Hence, it would be best suited for embedded applications.