Scope:  The recommended practice leverages Machine Learning (ML) in synthetic aperture applications, including radar, sonar, radiometry, magnetic resonance imaging, automotive radar, remote sensing, ultrasound, and other imaging modes. The recommended practice addresses the choice of Machine Learning architectures for different imaging scenarios and the appropriate training regime. The recommended practice provides techniques for object classification and segmentation in synthetic aperture images. The document explores the best data format for ML algorithms in different synthetic aperture applications. Examples of machine learning architectures described by the recommended practice include traditional approaches such as deep learning, autoencoders, reinforcement learning, transformer models, convolutional neural networks (CNNs), recurrent neural networks (RNNs), and new generative approaches. Further different data regimes are considered, including subsampled and sparse approaches, as well as fully sampled data acquisitions.