Google has released a dataset of 15,000 videos and 4 million images of everyday objects to advance researchers’ understanding of 3D object detection. The images of objects fall into one of nine categories, such as books or shoes, and are manually annotated to describe their position, orientation, and dimensions. Alongside the dataset, Google has also released a mobile 3D object detection application that uses machine learning to detect cameras, chairs, mugs, and shoes from 2-dimensional images and determine how they are positioned, such as a shoe laying flat.
Image: Google Objectron Dataset