Places: A 10 Million Image Database for Scene Recognition

The paper introduces the Places Database, motivated by how multi-million-item datasets let data-hungry machine learning algorithms approach near-human performance on tasks such as visual object and scene recognition. Places is a repository of 10 million scene photographs labeled with scene semantic categories, comprising a large and diverse list of the types of environments encountered in the world. Using state-of-the-art convolutional neural networks, the authors provide scene classification models, the Places-CNNs, as baselines trained on this data.

The Places-CNNs significantly outperform previous approaches to scene classification, and visualization of the trained networks shows that object detectors emerge as an intermediate representation of scene classification. With its high coverage and high diversity of exemplars, the Places Database along with the Places-CNNs offers a novel resource to guide future progress on scene recognition problems.