The advent of artificial intelligence and machine learning is influencing the manufacturing industryprofoundly, enabling unprecedented opportunities to improve manufacturing processes within the threedimensions time, quality and cost. With the introduction of digitization and industry 4.0, increasing amountsof data become available for processing and use in smart manufacturing systems. However, the various usecases for machine learning in manufacturing often require problem-specific datasets for training andevaluation of algorithms which are difficult to acquire, hindering both practitioners and academic researchersin this area. As the respective data frequently contains sensitive information, manufacturing companies rarelyrelease datasets to the public. Further, the relevant attributes and features of available datasets are usuallynot evident, requiring time-consuming analysis to evaluate if a dataset fits a given problem. As a result, itcan be challenging to develop and evaluate machine learning methods for manufacturing systems due to thelack of an overview of available datasets. This paper presents a comprehensive overview of 47 existing,publicly available datasets, mapped to various use cases in manufacturing with the goal of simplifying andstimulating research. The characteristics of the datasets are compared using a set of descriptive attributes toprovide an outline and guidance for further research and application of machine learning in manufacturing.In addition, suitable performance metrics for the evaluation of classification use cases in manufacturing arepresented.