Abstract

In recent years, applications such as Snapchat, Instagram, and Zoom have introduced perceptual manipulation filters to the masses. From a software perspective, filters are components that modify a video or audio stream. With this trend in mind, filters are expected to be ubiquitous in the future and a fascinating field for researchers. Since existing applications do not provide the degree of freedom and features required in research, it can be cumbersome to conduct research in this field. To solve this problem, this thesis conducts background research into filters and proposes a new conferencing platform. This platform is specifically designed for experiments with filters, gives researchers complete control over data and experiments, and enables them to use their own custom filters. To evaluate the platform, specific end-to-end video latency and API RTT testing tools were developed. A series of tests showed a stable average end-to-end video latency between 53.7ms and 81.2ms, depending on the contents of the input video and its dimensions. The API RTT averaged between 1.9ms and 15ms, depending on the packet size. Thanks to its expendability and comprehensive filter framework, the platform enables research with filters that were previously only feasible in complicated lab setups. Retesting results is also made easier thanks to the ability to share experiment setups and filters. Custom filters can be contributed to the platform since it is published as open source software.

Results/Implementation/Project Description

Conclusion

[ PDF (optional) ] 

[ Slides Kickoff/Final (optional)]