Bootstrapping Face Detection with Hard Negative Examples
This improves face detection accuracy for applications like security and photography, but is incremental as it builds on existing Faster R-CNN methods.
The paper tackles face detection by using hard negative mining to iteratively update a Faster R-CNN based detector, resulting in outperforming state-of-the-art detectors on the FDDB dataset.
Recently significant performance improvement in face detection was made possible by deeply trained convolutional networks. In this report, a novel approach for training state-of-the-art face detector is described. The key is to exploit the idea of hard negative mining and iteratively update the Faster R-CNN based face detector with the hard negatives harvested from a large set of background examples. We demonstrate that our face detector outperforms state-of-the-art detectors on the FDDB dataset, which is the de facto standard for evaluating face detection algorithms.