The focusing of a seismic image is directly linked to the accuracy of the velocity model. Therefore, a critical step in a seismic imaging workflow is to perform a focusing analysis on a seismic image to determine velocity errors. While the offset/aperture-angle axis is frequently used for focusing analysis, the physical (i.e., midpoint) axes of seismic images tend to be ignored as focusing analysis of geological structures is highly interpretative and difficult to automate. We have developed an automatic data-driven approach using convolutional neural networks to automate image-focusing analysis. Using focused and unfocused geological faults, we show that our method can make use of both spatial and offset/angle focusing information to robustly estimate velocity errors within seismic images. We demonstrate that our method correctly estimates velocity errors from a 2D Gulf of Mexico limited-aperture image where a traditional semblance-based approach fails. We also show that our method has the added benefit of improving the interpretation of faults within the image.