陈志波  (教授)

博士生导师

电子邮箱:

学位:博士

毕业院校:Tsinghua University

   

Extended HEVC Intra Coding (EHIC) Dataset

点击次数:

Introduction

Video encoder complexity reduction is a significant topic in the research area with a great impact on video industry. HEVC is the latest video coding standard and it achieves performance improvement under the cost of higher complexity. Deep learning can be an effective tool to decide the coding pattern in advance, which can significantly reduce the complexity. To meet the requirement of training data, we establish this EHIC dataset, which can serve for both the fast CU/PU size decision and fast intra-mode decision. This dataset includes the block with the coding pattern as training label.

Download

We are making the EHIC dataset available to the research community free of charge.

To get the publicly available release of the dataset, Please fill THIS FORM and the source code will be sent to you.

Database Description

The Extended HEVC Intra Coding (EHIC) Dataset is for both fast CU/PU size decision and intra-mode decision. Both high definition and low definition raw images are collected, corresponding to the resolution range of JCTVC standard test set. Then all the images are encoded by the HEVC reference software HM 16.9, while four QPs in {22, 27, 32, 37} are applied for encoding. Apart from the coding block, we provide the MNRC (minimal number of RDO candidates) label for fast intra-mode decision. And the traditional binary splitting label for fast size decision supplied. In addition, the RD-cost of each block is also included in this dataset, which can be used for the optimization of the training process.

Investigators

The investigators in this research are:

» Zhibo Chen (chenzhibo@ustc.edu.cn) Professor, Dept. of EEIS, USTC

» Jun Shi (shi1995@mail.ustc.edu.cn) Graduate student, Dept. of EEIS, USTC

» Weiping Li (wpli@ustc.edu.cn) Professor, Dept. of EEIS, USTC


上一条: Semantic Database Based on the Surveillance Scenarios

下一条: IMCL Stereoscopic Omnidirectional Image quality assessment Database (SOLID) - Phase II