One paper on ImageNet auto-annotation accepted to CVPR 2014. The camera-ready version is online. Object bounding-boxes output by this method are available