With the rapid construction of 5G infrastructure in China, the ultra HD video field has got perfect technical conditions for transmission and playback. Coupled with the strong demand for 4K program sources in the market, it is difficult for some open 4K Ultra HD channels to realize 24-hour no-repeated broadcasting. The main vedio contents of video-on-demand programs on video websites are still high-definition (720p, 1080p). Although many video websites have launched their own ultra HD channels, the number of ultra HD programs is limited.
GPU is good at large-scale parallel computing, so it is more suitable for deep learning than CPU. The computing power of a single GPU node is limited. To improve the video enhancement computing performance, more computing devices are needed to form a cluster and support distributed computing at the software level. The costs of software system maintenance and hardware procurement are very high.
Based on deep learning technology and training with massive data, the core algorithm of AI video enhancement solution can improve the resolution, noise reduction, HDR and frame rate of low-quality videos, so as to enhance the image quality of videos.
Based on the core algorithm, the software system can be used for product development, providing video storage, online playback and task management functions. It supports private deployment, public cloud and hybrid cloud deployment, which can be selected flexibly according to the customers' different needs, providing customers with "AI + cloud computing" solutions.
As the optimal solution to balance high processing capacity and cost, hybrid cloud deployment can separate data management from computing services, store and manage its own data at the lowest cost, and purchase cloud computing capacity according to the demand.