Alibaba Cloud released a real-time super-score tool on the side

Alibaba Cloud’s machine learning PAI team recently released a one-click end-to-end super-resolution tool, which can double the resolution of mobile video while keeping the device and network bandwidth unchanged. Can be enhanced to 1440p.

Super-resolution means "super-resolution", which refers to improving the clarity of low-resolution pictures or videos through deep learning to obtain higher resolutions, thereby achieving better viewing experience and recognition accuracy. By learning the known high-resolution resources through a special convolutional neural network (CNN), a model for super-resolution processing of low-resolution image and video resources can be realized.

This one-click end-to-end super-score tool equipped with the PAI-AISR model released by the Alibaba Cloud Machine Learning PAI team can perform accelerated optimizations such as quantitative pruning of the model, as well as multiple end-to-end devices (CPU / GPU / NPU). Heterogeneous computing and memory I/O optimization, etc.

According to reports, this super-resolution tool can enhance the video resolution to a maximum of 1440p on the mobile terminal, and at the same time, it can improve the aliasing and noise reduction, color enhancement, etc., support cross-platform use, and improve the viewing experience. Among them, the PAI-AISR algorithm requires a minimum of 50KB. At present, this technology has been applied in multiple apps such as Youku, Quark, and UC Browser.