This dataset is a customized subset of the Visual Wake Words (VWW) dataset, specifically optimized for downstream TinyML applications. Visual Wake Words represents a common micro-controller use case: determining whether a person is present in the frame (a "wake word" for vision applications). VWW dataset is entirely derived from the Microsoft COCO 2014/2017 dataset splits, filtered and re-labeled into a binary classification task. This dataset features a split-licensing structure inheriting from its upstream source, MS COCO: 1. Annotations: Any new or modified annotations, labels, and bounding boxes created for this project are released under the Creative Commons Attribution 4.0…