A Dataset for Machining Feature Detection in Rotation Symmetric Components

In Computer-Aided Process Planning (CAPP), machining feature recognition is the first and essential step for planning the manufacturing of the designed model. Due to the rapid variation in demand for products, it is desired to automate the process planning starting with detecting the features to be manufactured. As the ongoing researches with learning-based methods show promising results, the need for a suitable dataset for these methods arises. The existing datasets either lack the relation to the real use cases or the labeling is missing, which is critical for the learning-based methods. A synthetically generated labeled dataset is presented, that is built of real-like components. The features in samples of the dataset are selected from eleven manufacturing feature classes, which include features like thread and gear, that are not present in existing works. The proposed dataset also provides localization information by containing labeled faces. It enables transformation to various data representations, as it follows the ISO standard STEP 214 format. Four different complexity levels of the dataset are submitted so that the feature detection methods can be tested in different levels.