Modeling and analysis of complex flow behavior of impingement jets is a problem of significant importance in many engineering applications. Due to the nonlinear nature of these flows, traditional modeling methods often struggle to provide accurate representation of the flow features. Therefore, the goal of this work is to build a data-driven model for the available data to uncover the hidden features of the underlying dynamics, and to improve analysis and modeling of impinging jets. The available data consist of experimental velocimetry results of a circular impinging jet at a Reynolds number of 1260. The time-resolved particle image velocimetry (TR-PIV) technique was used to obtain velocity field data. An Autoencoder (AE) which is a special type of convolutional neural network is used for data compression and thus to learn the hidden features of the jet. The accuracy for reconstruction purposes was evaluated for various dimensions of the AE latent vector. According to the findings, the flow field image can be reconstructed using only one variable in the latent vector, which corresponds a reduction to 0.0015% in the size of the original flow image. The analysis of the spectral content of the AE variables revealed two primary frequency peaks, which coincided with those identified in the transverse velocity spectrum extracted from the main vortices' path. This suggests a connection between the AE variables and the vortical structures.