Stanford researchers develop algorithm to upgrade video streaming
As millions remain under lockdown and glued to their screens, researchers at Stanford University are working to make content streaming more seamless than ever. They have developed an algorithm that uses machine learning to reduce all the glitches and stalls people run into while streaming something on platforms like Netflix and Amazon Prime Video. Here's all about it.
Fugu, the machine learning tool
Dubbed as Fugu, the novel machine-learning algorithm strives to improve the video streaming experience by pushing only as much content as the viewer's internet connection can receive without degrading the quality. Now, this would be a major upgrade over current streaming systems, which struggle to adjust according to bandwidth, resulting in either unwanted hiccups or freakishly pixelated content that no one likes to watch.
Buffer-based algorithm used with most streaming systems
According to Stanford researchers, most streaming systems active these days rely on a Buffer-Based Algorithm (BBA), which checks how much video a device has in its buffer to decide whether it should be given high or low-quality content. For instance, if it learns that a device has less than 5 seconds of video, it will serve pixelated, low-quality content to avoid interruptions.
High quality video served when 15+ seconds in buffer
If the algorithm discovers that 15+ seconds of video is in the buffer, it serves high-quality video. Meanwhile, for anything between 5 and 15 seconds, it adjusts streaming quality accordingly.
Trial comparing Fugu with BBA, other algorithms
To test the effectiveness of the new Fugu algorithm, the Stanford team configured it to work on Puffer - a live TV service with over 100,000 users, along with four other streaming algorithms, including BBA. They monitored how each of the algorithms performed at handling the data streams and ensuring that users get optimum video quality with minimal disruptions.
Fugu outperformed all the algorithms
By the time the trial ended, Fugu had defeated all other algorithms by delivering the highest image resolution with minimal interruptions. The video quality processed by the algorithm was also more consistent compared to its counterparts. Moreover, the performance of the algorithm was so good that the viewers of Fugu-fed video streams continued to watch an average of 5-9% longer than other tested algorithms.
No word on integration yet
While it is evident that Fugu makes a better prospect for streaming services, there is no word on its integration yet. Also, before any of that, the researchers would have to carry out tests to verify the feasibility of the algorithm for a bigger viewer-base.
