Automatic Dataset Augmentation

TL;DR

This paper presents an automatic dataset augmentation method that leverages web data and deep learning to enlarge image datasets, improve object recognition accuracy, and reduce labeling costs.

Contribution

It introduces a novel approach combining web data and DCNNs to automatically label and augment datasets, enhancing recognition performance without manual labeling.

Findings

Automatically scales datasets from billions of web images

Significantly improves object recognition accuracy

Demonstrates effective use of web-derived supervisory information

Abstract

Large scale image dataset and deep convolutional neural network (DCNN) are two primary driving forces for the rapid progress made in generic object recognition tasks in recent years. While lots of network architectures have been continuously designed to pursue lower error rates, few efforts are devoted to enlarge existing datasets due to high labeling cost and unfair comparison issues. In this paper, we aim to achieve lower error rate by augmenting existing datasets in an automatic manner. Our method leverages both Web and DCNN, where Web provides massive images with rich contextual information, and DCNN replaces human to automatically label images under guidance of Web contextual information. Experiments show our method can automatically scale up existing datasets significantly from billions web pages with high accuracy, and significantly improve the performance on object recognition tasks by using the automatically augmented datasets, which demonstrates that more supervisory information has been automatically gathered from the Web. Both the dataset and models trained on the dataset are made publicly available.