Modeling and Evaluation of Tile-Based Adaptive 360-degree Video Streaming
360° video, spherical video, video streaming, ABR, adaptive transmission, dataset, transmission metrics.
virtual reality glasses on cellular devices. Given that humans perceive only a fraction of the visual sphere, transmitting the entire sphere results in an inefficient use of resources. In this new scenario, spatial segmentation of the video into tiles appears as a solution, accompanied by several new ABR (Adaptive Bit Rate) algorithms using DASH (Dynamic Adaptive Streaming over HTTP). However, evaluating the performance of these algorithms subjectively is a complex task, as each user interaction is unique. This work explores the creation and characterization of a database containing transmission and reproduction metrics, such as decoding time, bit rate and MSE (Mean Squared Error), allowing them to be used in conjunction with head movement recording to evaluate different algorithms under similar conditions. To this end, in addition to the metrics database, a database recording the tiles viewed by each user was developed, together with a Python module for geometric manipulation of the spherical video. Based on these data, several techniques are presented for analyzing and evaluating the emulation of adaptive transmission of 360° videos with tiles