Fraunhofer FOKUS, FAME | Projects WS 2020/2021

Projects in WS 2020/2021

Web App with advanced statistics and predictions for sports fantasy
Real-Time Streaming Analytics UI for CMCD and SAND
Dash.js
Deep Encode
Immersive Video Conferencing

Web App with advanced statistics and predictions for sports fantasy

So called fantasy games are getting more and more popular with sports fans. They usually come with advanced statistics to help players make decisions, e.g. who to line up, buy, sell etc. Sometimes these statistics are limited or lack certain features (e.g. predictions). As part of this project, you will develop additional helper tools for a fantasy game.

Your tasks:

Set-up a dataset with fantasy sports data
Aggregate/transform the data
Build a Web App to visualize the data
Predict data based on historical data and third-party sources

Required skills:

Web development
Enthusiastic sport fan

Related links:

Supervisor: Stefan Pham

Real-Time Streaming Analytics UI for CMCD and SAND

Real-Time streaming analytics enable content providers to identify problems on certain platforms or devices. Specifications like Server and Network assisted DASH (SAND) and Common-Media-Client-Data(CMCD) define the format and the types of metrics send from a client to a metric server. A user interface build on top of the collected data helps evaluating and structuring the massive amount of information.

Your tasks:

Understand specifications: DASH, SAND and CMCD
Summarize metrics
Build a real-time analytics UI, e.g. using Grafana/Kibana

Required skills:

Basic understanding and interest in media streaming
JavaScript, DB

Related Links:

Supervisor: Stefan Pham

Dash.js

dash.js is a free, open source MPEG-DASH player that serves as a JavaScript reference client for implementing production grade DASH players.The dash.js player is written in JavaScript and relies on the Media Source Extensions (MSE) and the Encrypted Media Extensions (EME). Since a majority of the common target platforms like Desktop browsers, mobile devices, Smart TVs and gaming consoles support both MSE and EME, dash.js is a prominent option to be used as a foundation for a production grade player.

The goal of this project is to add a new reference UI that allows the configuration of the various dash.js settings. The existing reference UI can be found here: http://reference.dashif.org/dash.js/nightly/samples/dash-if-reference-player/index.html.

Your tasks:

Understand the principles behind adaptive streaming
Get familiar with the dash.js API
Examine the existing reference UI and its features
Implement a new reference UI supporting all of the parameters from Settings.js
Use an up-to-date framework like React.js or Vue.js

Required skills:

Interest in media streaming
Knowledge in web development
Ideally: Skills in web design

Related Links:

Supervisor: Daniel Silhavy

Deep Encode

Video streaming content differs in terms of complexity and requires title-specific encoding settings to achieve a certain visual quality. With per-title encoding, however, several test encodes are needed, which results in a high amount of storage and bitrates used. The Deep Encode project utilizes machine learning models and provides encoding setting predictions in order to avoid the computationally heavy test encodes.

Your tasks:

Understand the basic principles of encoding and per-title encoding
Develop an evaluation framework that compares the performances between different machine learning models
How can the machine learning models be improved in order to enhance video quality predictions?

Required skills:

Interest in video encoding and media streaming
Familiarity with machine learning
Basic Python or R skills

Related links:

Video encoding/streaming overview: https://github.com/leandromoreira/digital_video_introduction
Per-title encoding: https://websites.fraunhofer.de/video-dev/per-title-encoding/
https://netflixtechblog.com/per-title-encode-optimization-7e99442b62a2
https://ieeexplore.ieee.org/document/7532605
VMAF: https://streaminglearningcenter.com/blogs/collection-of-vmaf-resources.html

Supervisors: Anita Chen, Christoph Müller

Immersive Video Conferencing

This project is about evaluating and developing enhanced video conferencing application based on open source frameworks and Web technologies. Different approaches for video communication should be explored, developed and evaluated. The target applications are intended to run in Web browsers on desktop and mobile. Therefore, Web technologies like WebRTC and other HTML5 related APIs need to be considered as a foundation for the video conferencing application.

Furthermore, the video conferencing tools should easily be integrated into learning management systems such as Moodle (at TU known as ISIS) or ILIAS and linked to given learning media. Therefore, open standards and specifications for learning environments should be used (e.g. Learning Tools Interoperability [LTI] or computer-managed instruction [cmi5]).

The list of tasks below describes dedicated features that can be implemented in different groups. Each group will work only on 1-2 tasks and not on all of them.

Your tasks:

Develop Multiparty Video Conferencing Application based on open source (Web) technologies and frameworks
Integration of Video Conferencing Tools in learning management systems (LMS) as well as linking the learning media with the video conference (e.g., in terms of presentation slides etc.)
Explore and evaluate different interfaces for the integration of video services in LMS
Explore and evaluate different topologies like Mesh, MCU (Multipoint Conferencing Unit) and SFU (Selective Forwarding Unit)
Explore and evaluate different visualisation modes (Grid, VR/AR, Hybrid, ...) of participants while taking device capabilities into consideration (device type, screen size, ...)
Explore and evaluate multiscreen and multi-device capabilities by allowing to distribute the video conference of multiple devices/screens (Example: connect mobile browser to desktop browser or TV and split the view on multiple screens).
Explore concepts for better organisation of video conferences and develop a prototype (can be a mockup) to evaluate the approach (Examples: Integrate Meeting Agenda, Time boxing, auto-generate minutes, better coordination across participants in large meetings, )
Explore and evaluate AI techniques to enhance video conferencing experience while maintaining privacy (Examples: Virtual Backgrounds, Video filters, Subtitles from voice, Noise Reduction)

Required skills: