Deep Render Claims Huge Boost in Video Compression
A start-up by scientists from Imperial College, London and other universities, has claimed that it can improve the compression efficiency of current codecs by 5X using AI-based techniques, but believes that it can achieve ‘up to’ 50X reductions in file size eventually. The firm’s software-based codec is said to be very flexible and ‘fine-grained’.
Of course, better codecs always help higher resolution and higher quality content such as 8K, but it’s unclear from the firm at the moment whether there is a big overhead in the computational power needed for each of the pipeline (the typical trade-off in more sophisticated codecs). It is also not clear to what extent the processes can be implemented in optimized hardware. If they need large general purpose processors for decompression, that could be a barrier to widespread use.
The company earlier this year raised $9 million in funding and has received EU grants.
