EndlessCL-Simulator-Dataset for experimentation and reproducibility of the empirical investigation in our paper - A Procedural World Generation Framework for Systematic Evaluation of Continual Learning - Timm Hess, Martin Mundt, Iuliia Pliushch, Visvanathan Ramesh, 2021. For ease of use, we provide the dataset separately as video sequences (successive sequence of video frames with semantic object class annotation) and image-patch classification collection. Each dataset contains train and test data. Files contained: IncrementalClass_Video: Representing the most commonly investigated continual scenario, our video stream consists of four video sub-sequences, each adding one distinct object class(tree, car, people, streetlamp) to the task. The first task additionally provides examples for the 'background' category, which contains a combination of categories being part of every sub-sequence, i.e. street, sidewalks, terrain, sky, and buildings. IncrementalClasses_Classification: Image patch classification dataset based on the above video sequence. IncrementalLighting_Video: The incremental lighting video sequence is base on a progressive decrease in illumination intensity as a single generative factor, without adjustments to the illumination color. Objects of all categories are present in all sub-sequences. IncrementalLighting_Classification: Image patch classification dataset based on the above video sequence. IncrementalWeather_Video: The incremental weather stream is a compilation of distinct weather conditions (clear day, fog, rain, snow, overcast) that are iteratively explored in sub-sequences. IncrementalWeather_Classification: Image patch classification dataset based on the above video sequence. Please see the README, inside the zip file, for instructions. The standalone simulator executable is available at: https://doi.org/10.5281/zenodo.4899294 The underlying source code of this simulator executable is made available at: https://github.com/ccc-frankfurt/EndlessCL-Simulator-Source