Offline 1000-Class Classification on a Smartphone

Yoshiyuki Kawano, Keiji Yanai
2014 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops  
In this demo, we propose an offline large-scale image classification system on a smartphone. The proposed system can classify 1000-class objects in the ILSVRC2012 dataset in 0.270 seconds. To implement a 1000-class object classification system, we compress the weight vectors of linear classifiers, which leads only slight performance loss.
doi:10.1109/cvprw.2014.35 dblp:conf/cvpr/KawanoY14 fatcat:sgsjseqdgbgszayepc6l2qwvla