Near-Lossless Deep Feature Compression for Collaborative Intelligence

TL;DR

This paper introduces a near-lossless deep feature compression method for collaborative intelligence, reducing data transmission rates and enabling image reconstruction from compressed features, thereby improving efficiency in mobile-cloud neural network deployment.

Contribution

It proposes a simple, effective near-lossless compressor for deep features and a method to reconstruct input images from compressed features, enhancing collaborative intelligence systems.

Findings

Achieves up to 5% bit rate reduction over HEVC-Intra

Outperforms other popular image codecs in compression efficiency

Enables image reconstruction from compressed deep features

Abstract

Collaborative intelligence is a new paradigm for efficient deployment of deep neural networks across the mobile-cloud infrastructure. By dividing the network between the mobile and the cloud, it is possible to distribute the computational workload such that the overall energy and/or latency of the system is minimized. However, this necessitates sending deep feature data from the mobile to the cloud in order to perform inference. In this work, we examine the differences between the deep feature data and natural image data, and propose a simple and effective near-lossless deep feature compressor. The proposed method achieves up to 5% bit rate reduction compared to HEVC-Intra and even more against other popular image codecs. Finally, we suggest an approach for reconstructing the input image from compressed deep features in the cloud, that could serve to supplement the inference performed by the deep model.