Last update：2022-04-13 14:59:13
With deep learning breakthroughs and the technological and policy supports, Artificial Intelligence era is right around the corner. Upholding the principle of technology-driven product development, CDNetworks rolls out a series of Intelligent Video products, where AI technology empowers CDN and the AI technology go-live scheme is explored. Resolution rebuild is a critical feature for the CDNetworks Intelligent Video products.
The explosive growth of Internet video business in the last two years has entailed the increasingly fierce competition in fields such as live streaming and short video business. Video products are in pursuit of a higher and higher resolution, but this pursuit is constrained by traffic costs. Thus, the applications of short videos and live streaming are mainly in low resolution, which compromises the viewing experience and the competitiveness.
Under such circumstances, the video business gains more momentum from resolution rebuild, which is to reset the resolution from low to high in the terminal based on deep learning algorithm. Compared with the traditional interpolation algorithm, the deep learning algorithm model is more meticulous in image detail processing, so there will be no jags and burrs generated and the image display performance will be upgraded obviously. With the deployment of resolution rebuild model in the terminal, the conversion from low resolution to high resolution of video will be conducted in the viewing terminal, and users will enjoy a HD video of richer images showing more details without increasing bandwidth costs for ISPs.
This feature is suitable for the mobile live streaming and short form video applications, as their video resolutions are low, mainly under 360P. By upgrading the resolution from 360P to 720P or higher, the user viewing experience will be enhanced greatly.
By using the image resolution rebuild feature, customers owning a great number of old video resources can repair and reset those video resources.
Resolution rebuild feature will upgrade video quality significantly, which can be widely applied in security identification, drone, intelligent manufacturing video analysis and so on.
The use of resolution rebuild requires SDK embedding. After the video is decoded by the player, input the video frame to SDK, and SDK will output a high-resolution video frame on the basis of the input frame information and the pre-loaded deep learning model. For now, the SDK is available in a basic and an advanced version for IOS and Android terminals.
|Supported feature||Image quality repair, 2 times super-resolution and 3 times super-resolution|
|Supported video formats||yuv frame input, yuv frame output|
|Supported video resolution input||360p maximum for mobile end|
|Supported devices||For Android terminals: support devices using the Qualcomm Snapdragon 660, 820, 821, 835 or 845 For iOS terminals: support devices later than iPhone 6s/iPad air 2|
For iOS terminals: the Bundle ID of iOS APP is needed, and it will be used as the unique identifier.
For Android terminals: the application package name such as com.test.sdk is needed.
Select the terminal on the basis of the demand, obtain the unique identifier, and feedback the information to customer service engineers for test enabling.