Software
MCSL actively contributes open source software to the video coding standards and deep learning.
People who wish to access the software request the software coordinator for report access.
We provide natural multi view video datasets named “SKKU_VRroom1D” and “SKKU_VRroom2D” which consists of 30 cameras and 300 frames.
You can download the dataset in link below.
High-efficiency video coding (HEVC) tile merger is an application for merging tile bitstreams to generate a single bitstream. It is based on HEVC test model (HM) 16.20, and fully written in C++.
Coordinator:
Jong-Beom Jeong (uof4949@skku.edu)
Kvazaar-compliant high-efficiency video coding (HEVC) tile extractor is an application for extracting a tile from a motion-constrained tile set (MCTS) bitstream. Unlike the existing HEVC tile extractor, it uses slice segment address to find the target slice, thereby it can be used for Kvazaar MCTS bitstream. It is based on HEVC test model (HM) 16.20, and fully written in C++.
Coordinator:
Jong-Beom Jeong (uof4949@skku.edu)
The acoustic feature of climax part is used and analyzed to generate animated GIFs for music videos. It is based on Keras, TensorFlow, FFmpeg, and fully written in Python.
Coordinator:
Ghulam Mujtaba (mujtaba@gachon.ac.kr)
High-efficiency video coding (HEVC) sub-picture extractor is an application for extracting a sub-picture which contains multiple tiles from a motion-constrained tile set (MCTS) bitstream. It is based on HEVC test model (HM) 16.20, and fully written in C++.
Coordinator:
Jong-Beom Jeong (uof4949@skku.edu)
High-efficiency video coding (HEVC) multiple-tile extractor (MTE) is an application for extracting multiple viewport tiles from a motion-constrained tile set (MCTS) bitstream and generating single bitstream which has the same resolution with the input bitstream. It is based on HEVC test model (HM) 16.20, and fully written in C++.
Coordinator:
Jong-Beom Jeong (uof4949@skku.edu)
Versatile video coding (VVC) tile extractor is an application for extracting tile bitstream from a motion-constrained tile set (MCTS) bitstream. It is based on VVC test model (VTM) 7.3, and fully written in C++.
Coordinator:
Jong-Beom Jeong (uof4949@skku.edu)
6DoF 360-degree video viewport tile selector (VTS) is an application for selecting viewport area tiles of multiple 360-degree videos. It is based on test model for immersive video (TMIV) v1.0, and fully written in C++.
Coordinator:
Jong-Beom Jeong (uof4949@skku.edu)
3DoF+ 360-degree video pruner is an application for removing the correlation between multiple 360-degree videos using view synthesis. It removes the pixels from sparse views which a central view already have. It is based on reference view synthesizer (RVS) 2.0.1, and fully written in C++. This SW was supported by a reference SW development project funded by LG Electronics Research. After the project, MCSL improved the result SW, and uploaded it to gitlab private repository. People who wish to access may request the software coordinator.
Coordinator:
Jong-Beom Jeong (uof4949@skku.edu)
Soonbin Lee (soonbinlee@skku.edu)
3DoF+ 360-degree video packer is an application for extracting the rectangular patches which contain residuals from the sparse views using region growing and merging them into a packed view using stride. It is fully written in C++. This SW was supported by a reference SW development project funded by LG Electronics Research. After the project, MCSL improved the result SW, and uploaded it to gitlab private repository. People who wish to access may request the software coordinator.
Coordinator:
Jong-Beom Jeong (uof4949@skku.edu)
Soonbin Lee (soonbinlee@skku.edu)
3DoF+ 360-degree video unpacker is an application for extracting the rectangular patches from a packed view and reconstructing the sparse views. It is fully written in C++. This SW was supported by a reference SW development project funded by LG Electronics Research. After the project, MCSL improved the result SW, and uploaded it to gitlab private repository. People who wish to access may request the software coordinator.
Coordinator:
Jong-Beom Jeong (uof4949@skku.edu)
Soonbin Lee (soonbinlee@skku.edu)
3DoF+ 360-degree video reconstructor is an application for reconstructing the removed correlations between sparse views and a central view using view synthesis. It is based on reference view synthesizer (RVS) 2.0.1, and fully written in C++. This SW was supported by a reference SW development project funded by LG Electronics Research. After the project, MCSL improved the result SW, and uploaded it to gitlab private repository. People who wish to access may request the software coordinator.
Coordinator:
Jong-Beom Jeong (uof4949@skku.edu)
Soonbin Lee (soonbinlee@skku.edu)
Image web crawler + Object detection(YOLO V3) aims to reduce the noise of existing web crawl results, and written in C++, Python 3.
Coordinator:
Eun-Soo Park (espark804@skku.edu)