Developing an emulator for 360° video : intended for algorithm development
Abstract: Streaming 360° video has become more commonplace with content delivery services such as YouTube having support for it. By its nature, 360° video requires more bandwidth as only a fraction of the image is actually in view, while the user is expecting the same "in view" quality as with a regular video. Several studies and lots of work have been done to mitigate this higher demand for bandwidth. One solution is advanced algorithms that take in to account the direction that the user is looking when fetching the video from the server; e.g., by fetching content that is not in the user’s view at a lower quality or by not fetching this data at all. Developing these algorithms is a timely process, especially in the later stages where tweaking one parameter might require the video to be re-encoded, and therefore taking up time that could otherwise be spent on getting results and continued iteration on the algorithm. The viewer should also be considered as the best experience might not correlate with the mathematically best solution calculated by the algorithm. This thesis presents a modular emulator that allows for easy implementation of fetching algorithms that make use of state-of-the-art techniques. It intends to reduce the time it takes to iterate over an algorithm through removing the need to set up a server and encode the video in all of the wanted quality levels when a parameter change would require it. It also makes it easy to include the viewer in the process so that the subjective performance is taken into consideration. The emulator is evaluated through the implementation and evaluation of two algorithms, one serving as a baseline to the second one, which is based on an algorithm developed by another group of researchers. These algorithms are tested on two different types of 360° videos, under four different network conditions and with two values for the maximum buffer size. The results from the evaluation of the two algorithms suggest that the emulator functions as intended from a technical point of view, and as such fulfills its purpose. There is, however, future work that would further prove the emulators performance in regards to replicating real scenarios and a few examples are suggested.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)