iSIZE's BitSave preprocessor solution aims to reduce the size of video streams and so reduce the carbon footprint associated with video streaming. At the input to the encoder, BitSave takes in raw video and outputs it with a bandwidth saving of about 20% on top of the existing encoding gains. Using iSIZE’s deep neural network technology, BitSave reduces encoding complexity, saving data centre processing power and energy consumption.
The infrastructure for distributing online and mobile video is well established and largely unchangeable; upgrading codecs takes time and deployment risks are high. With the surge in internet video traffic, there’s a huge environmental impact that comes from the growth in streaming services. iSIZE's BitSave preprocessor solution aims to reduce the size of video streams and so reduce the carbon footprint associated with video streaming. BitSave sits at the input to the encoder, taking in raw video and outputting it with a bandwidth saving of around 20% on top of the existing encoding gains.
No changes are required in the encoding, stream packaging, streaming and decoding. Using iSIZE’s deep neural network technology, BitSave reduces encoding complexity, saving data centre processing power and energy consumption.BitSave is being continually developed and within the last 12 months was shown to allow for nearly 15% average bitrate reduction over state-of-the-art AVC and AV1 encoders. This is a drastic improvement given that obtaining even 5% average saving over the state-of-the-art is a very complex R&D endeavour that tends to take more than 24 months to be achieved. (as published in joint whitepaper with Intel) Due to drastic reduction in complexity of the inference architecture of BitSave, this is now achieved at extremely high speed (over 200 frames-per-second for 1080p video) on mainstream Intel, AMD and NVIDIA hardware.
This makes BitSave widely applicable as a preprocessor for both lightweight legacy encoders like AVC, as well as more advanced encoders like VP9 and AV1, in both software and hardware. This was also recognised in the last 12 months with publications of papers at the top-tier IEEE International Conference on Computer Vision and Pattern Recognition (CVPR’21), the SMPTE Annual Technical Conference (ATC - to take place in November 2021, as well by accepting iSIZE as a member of the AOMEDIA consortium that standardizes AV1, AV2 and beyond. The solution also forms key element of a joint project, undertaken in partnership with BBC R&D and Queen Mary University of London that received an $1m research grant from Innovate UK. The improvements of BitSave and the recent commercial uptake were the primary reason iSIZE was able to raise a $6.5m funding round led by Octopus Ventures.