In this paper, we compare and evaluate the effectiveness of four deep learning models for sequential data—Vanilla Artificial Neural Networks (ANNs), 1-dimensional Convolutional Neural Networks (CNNs), Gated Recurrent Unit (GRU)-based RNNs, and Long Short Term Memory (LSTM)-based RNNs. The performance of these networks in learning various features of interest in our data are compared with each other. We apply these deep learning algorithms to the simple trajectory prediction problem of estimating Eulerian kinematic properties of motion of a flying solid body in two dimensions. We report how the various models compare against each other and show that LSTMs outperform the others. We also show an interesting result-that the number of tracked points is immaterial in this inferencing process. This may go against the intuition that the accuracy of prediction and quick convergence of a model would depend on the number of visual keypoints tracked. © 2019, American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved.