2C-Net: integrate image compression and classification via deep neural network

Providing effective support for intelligent vision tasks without image reconstruction can save numerous computational costs in the era of big data. With help Deep Neural Network (DNN), integrating compression and at a feature representation level becomes new promising approach. But how to perform non-linear transformation extract patterns simultaneously within shared DNN remains an open problem. In this paper, versatile framework is studied explore common representations both classification. A fully latent extracted more compact way classification task. The General Feature Extraction Feature-Analytic Classifier are proposed generate utilize representation. Then, whole joint optimized by considering multiple factors (i.e., rate, quality, accuracy). Extensive experiments carried out validate that proposals improve performance learning-based results show method outperforms conventional codecs like BPG JPEG2000 efficiency, while achieving acceptable accuracy on different datasets reconstruction.