Search

This paper presents a video coding scheme that combines traditional optimization methods with deep learning methods based on the Enhanced Compression Model (ECM). In this paper, the traditional optimization methods adaptively adjust the quantization parameter (QP). The key frame QP offset is set according to the video content characteristics, and the coding tree unit (CTU) level QP of all frames is also adjusted according to the spatial-temporal perception information. Block importance mapping technology (BIM) is also introduced, which adjusts the QP according to the block importance. Meanwhile, the deep learning methods propose a convolutional neural network-based loop filter (CNNLF), which is turned on/off based on the rate-distortion optimization at the CTU and frame level. Besides, intra-prediction using neural networks (NN-intra) is proposed to further improve compression quality, where 8 neural networks are used for predicting blocks of different sizes. The experimental results show that compared with ECM-3.0, the proposed traditional methods and adding deep learning methods improve the PSNR by 0.54 dB and 1 dB at 0.05Mbps, respectively; 0.38 dB and 0.71dB at 0.5 Mbps, respectively, which proves the superiority of our method.

In a traditional cloud storage system, users benefit from the convenience it provides but also take the risk of certain security and privacy issues. To ensure confidentiality while maintaining data sharing capabilities, the Ciphertext-Policy Attribute-based Encryption (CP-ABE) scheme can be used to achieve fine-grained access control in cloud services. However, existing approaches are impaired by three critical concerns: illegal authorization, key disclosure, and privacy leakage. To address these, we propose a blockchain-based data governance system that employs blockchain technology and attribute-based encryption to prevent privacy leakage and credential misuse. First, our ABE encryption system can handle multi-authority use cases while protecting identity privacy and hiding access policy, which also protects data sharing against corrupt authorities. Second, applying the Advanced Encryption Standard (AES) for data encryption makes the whole system efficient and responsive to real-world conditions. Furthermore, the encrypted data is stored in a decentralized storage system such as IPFS, which does not rely on any centralized service provider and is, therefore, resilient against single-point failures. Third, illegal authorization activity can be readily identified through the logged on-chain data. Besides the system design, we also provide security proofs to demonstrate the robustness of the proposed system., Comment: 23 pages, 19 algorithms, 1 figure

Blockchain technology enables secure tokens transfers in digital marketplaces, and recent advances in this field provide other desirable properties such as efficiency, privacy, and price stability. However, these properties do not always generalize to a setting across multiple independent blockchains. Despite the growing number of existing blockchain platforms, there is a lack of an overarching framework whose components provide all of the necessary properties for practical cross-chain commerce. We present BlockChain I/O to provide such a framework. BlockChain I/O introduces entities called cross-chain services to relay information between different blockchains. The proposed design ensures that cross-chain services cannot violate transaction safety, and they are furthermore disincentivized from other types of misbehavior through an audit system. BlockChain I/O uses native stablecoins to mitigate price fluctuations, and a decentralized ID system to allow users to prove aspects of their identity without violating privacy. After presenting the core architecture of BlockChain I/O, we demonstrate how to use it to implement a cross-chain marketplace and discuss how its desirable properties continue to hold in the end-to-end system. Finally, we use experimental evaluations to demonstrate BlockChain I/O's practical performance.

A plethora of different blockchain platforms have emerged in recent years, but many of them operate in silos. As such, there is a need for reliable cross-chain communication to enable blockchain interoperability. Blockchain interoperability is challenging because transactions can typically not be reverted - as such, if one transaction is committed then the protocol must ensure that all related transactions are committed as well. Existing interoperability approaches, e.g., Cosmos and Polkadot, are limited in the sense that they only support interoperability between their own subchains, or require intrusive changes to existing blockchains. To overcome this limitation, we propose PIEChain, a general, Kafka-based cross-chain communication framework. We utilize PIEChain for a practical case study: a cross-chain auction in which users who hold tokens on multiple chains bid for a ticket sold on another chain. PIEChain is the first publicly available, practical implementation of a general framework for cross-chain communication.

Ensemble learning has gain attention in resent deep learning research as a way to further boost the accuracy and generalizability of deep neural network (DNN) models. Recent ensemble training method explores different training algorithms or settings on multiple sub-models with the same model architecture, which lead to significant burden on memory and computation cost of the ensemble model. Meanwhile, the heurtsically induced diversity may not lead to significant performance gain. We propose a new prespective on exploring the intrinsic diversity within a model architecture to build efficient DNN ensemble. We make an intriguing observation that pruning and quantization, while both leading to efficient model architecture at the cost of small accuracy drop, leads to distinct behavior in the decision boundary. To this end, we propose Heterogeneously Compressed Ensemble (HCE), where we build an efficient ensemble with the pruned and quantized variants from a pretrained DNN model. An diversity-aware training objective is proposed to further boost the performance of the HCE ensemble. Experiemnt result shows that HCE achieves significant improvement in the efficiency-accuracy tradeoff comparing to both traditional DNN ensemble training methods and previous model compression methods.

Recently, deep neural networks (DNN) have been widely used in speaker recognition area. In order to achieve fast response time and high accuracy, the requirements for hardware resources increase rapidly. However, as the speaker recognition application is often implemented on mobile devices, it is necessary to maintain a low computational cost while keeping high accuracy in far-field condition. In this paper, we apply structural sparsification on time-delay neural networks (TDNN) to remove redundant structures and accelerate the execution. On our targeted hardware, our model can remove 60% of parameters and only slightly increasing equal error rate (EER) by 0.18% while our structural sparse model can achieve more than 1.5x speedup., Comment: submitted to icassp2020

Computer vision has achieved impressive progress in recent years. Meanwhile, mobile phones have become the primary computing platforms for millions of people. In addition to mobile phones, many autonomous systems rely on visual data for making decisions and some of these systems have limited energy (such as unmanned aerial vehicles also called drones and mobile robots). These systems rely on batteries and energy efficiency is critical. This article serves two main purposes: (1) Examine the state-of-the-art for low-power solutions to detect objects in images. Since 2015, the IEEE Annual International Low-Power Image Recognition Challenge (LPIRC) has been held to identify the most energy-efficient computer vision solutions. This article summarizes 2018 winners' solutions. (2) Suggest directions for research as well as opportunities for low-power computer vision., Comment: Preprint, Accepted by IEEE Journal on Emerging and Selected Topics in Circuits and Systems. arXiv admin note: substantial text overlap with arXiv:1810.01732

The Low-Power Image Recognition Challenge (LPIRC, https://rebootingcomputing.ieee.org/lpirc) is an annual competition started in 2015. The competition identifies the best technologies that can classify and detect objects in images efficiently (short execution time and low energy consumption) and accurately (high precision). Over the four years, the winners' scores have improved more than 24 times. As computer vision is widely used in many battery-powered systems (such as drones and mobile phones), the need for low-power computer vision will become increasingly important. This paper summarizes LPIRC 2018 by describing the three different tracks and the winners' solutions., Comment: 13 pages, workshop in 2018 CVPR, competition, low-power, image recognition