This article presents the data collection process for the classification of partial discharges in electrical generators using PNG format images. The data were collected through field measurements on over 40 generators in various locations in Colombia, in addition to utilizing a partial discharge simulator provided by Omicron Energy. Throughout the collection process, special attention was given to the accuracy and coherence of the images, avoiding deformations and distortions that could impact the nature of partial discharges. Emphasis was placed on achieving high resolution in phase-resolved patterns (PRPD) to effectively correlate them with the adjacent physical phenomenon. The analysis focused on classifying the images according to the type of partial discharge, identifying them as internal, surface, or corona discharges. The obtained pulse patterns are represented in RGB color, which aids in assessing the repeatability of pulses across their distribution. These data hold potential for the development of pattern classification software for generator monitoring systems. They enable the training and validation of classification algorithms, simplifying the automated detection and analysis of partial discharges in electrical generators. Their applicability extends beyond the electrical industry and can be valuable in other fields requiring complex signal and pattern analysis. The article highlights the rigorous data collection process and precise analysis conducted to obtain a valuable set of PNG format images for partial discharge classification. These data have significant potential in advancing pattern classification software, driving progress in the monitoring and analysis of electrical generators.