Super-resolution reconstruction method and apparatus for adaptive streaming media and server

1 . A super-resolution reconstruction method for an adaptive streaming of media over a network, the super-resolution reconstruction method comprising:

streaming an instance of a streamable media to a terminal device via the network;

identifying a streaming resolution of the instance of the streamable media;

based on the streaming resolution being less than a first target super-resolution, acquiring, by the terminal device from a server of the network, a first target super-resolution neural network model corresponding to the streaming resolution and the first target super-resolution;

reconstructing the instance of the streamable media streamed to the terminal device into an instance of the streamable media at the first target super-resolution using the first target super-resolution neural network model;

playing the reconstructed instance of the streamable media by the terminal device,

wherein the first target super-resolution neural network model is obtained by performing model training in advance by taking a media at the streaming resolution received from the server as an input and taking a media at the first target super-resolution received from the server as a learning objective,

wherein the super-resolution reconstruction method further comprises:

generating a plurality of target super-resolution neural network models of the streamable media; and

storing the plurality of target super-resolution neural network models on the server, and

wherein the acquiring, by the terminal device, the first target super-resolution neural network model from the server further comprises:

sending, by the terminal device, a super-resolution neural network acquisition request to the server via the network, wherein the super-resolution neural network acquisition request comprises the streaming resolution and a requested target super-resolution from among a plurality of target super-resolutions; and

receiving, by the terminal device, the respective target super-resolution neural network model from among the plurality of target super-resolution neural network models corresponding to a first resolution and the requested target super-resolution.

2 . The super-resolution reconstruction method of claim 1 , further comprising:

generating the first target super-resolution neural network model of the streamable media; and

storing the first target super-resolution neural network model on the server.

3 . The method of claim 1 ,

wherein each respective target super-resolution neural network model of the plurality of target super-resolution neural network models corresponds to the first resolution and a respective target super-resolution of the plurality of target super-resolutions, and

wherein the plurality of target super-resolutions includes the first target super-resolution, the plurality of target super-resolution neural network models includes the first target super-resolution neural network model, and the first resolution is lower than each respective target super-resolution of the plurality of target super-resolutions.

4 . The super-resolution reconstruction method of claim 1 , wherein the first target super-resolution neural network model is encoded with an instance of the streamable media at the streaming resolution.

5 . The super-resolution reconstruction method of claim 1 , further comprising:

identifying, by the terminal device, a user type; and

identifying, based on the identified user type, the first target super-resolution.

6 . The super-resolution reconstruction method of claim 5 , wherein the identifying, based on the identified user type, further comprises:

based on identifying the user type as an ordinary-level user, identifying the first target super-resolution to be a target resolution currently selected by a user, and

based on identifying the user type as an advanced-level user, identifying the first target super-resolution to be a preset advanced playing resolution.

7 . A super-resolution reconstruction terminal device for adaptive streaming of media comprising:

at least one memory configured to store at least one instruction; and

at least one processor configured to execute the at least one instruction,

wherein the at least one instruction, when executed by the at least one processor, causes the super-resolution reconstruction terminal device to:

receive a streamed instance of a streamable media;

identify a streaming resolution of the streamed instance of the streamable media;

based on the streaming resolution being less than a first target super-resolution, acquire from a server a first target super-resolution neural network model corresponding to the streaming resolution and the first target super-resolution;

reconstruct the streamed instance of the streamable media into an instance of the streamable media at the first target super-resolution using the first target super-resolution neural network model; and

play the reconstructed instance of the streamable media,

wherein the first target super-resolution neural network model is obtained by performing model training in advance by taking a media at the streaming resolution received from the server as an input and taking a media at the first target super-resolution received from the server as a learning objective, and

wherein the at least one instruction, when executed by the at least one processor, further causes the super-resolution reconstruction terminal device to:

send to the server a super-resolution neural network acquisition request comprising the streaming resolution and a requested target super-resolution from among a plurality of target super-resolutions, and

receive a target super-resolution neural network model, from among a plurality of target super-resolution neural network models, corresponding to the streaming resolution and the requested target super-resolution.

8 . The super-resolution reconstruction terminal device of claim 7 , wherein each respective target super-resolution neural network model of the plurality of target super-resolution neural network models corresponds to a first resolution and a respective target super-resolution of the plurality of target super-resolutions, and

wherein the plurality of target super-resolutions includes the first target super-resolution, the plurality of target super-resolution neural network models includes the first target super-resolution neural network model, and the first resolution is lower than each respective target super-resolution of the plurality of target super-resolutions.

9 . The super-resolution reconstruction terminal device of claim 7 , wherein the at least one instruction, when executed by the at least one processor, further causes the super-resolution reconstruction terminal device to:

identify a user type; and

identify, based on the identified user type, the first target super-resolution.

10 . The super-resolution reconstruction terminal device of claim 9 , wherein the at least one instruction, when executed by the at least one processor, further causes the super-resolution reconstruction terminal device to:

based on the user type being identified to be an ordinary-level user, identify the first target super-resolution to be a target resolution currently selected by a user, and

based on the user type being identified to be an advanced-level user, identify the first target super-resolution to be a preset advanced playing resolution.

11 . A super-resolution reconstruction system for adaptive streaming of media over a network, the super-resolution reconstruction system comprising:

a server comprising at least one server memory configured to store at least one server instruction, and at least one server processor configured to execute the at least one server instruction; and

a terminal device comprising at least one terminal device memory configured to store at least one terminal device instruction, and at least one terminal device processor configured to execute the at least one terminal device instruction,

wherein the at least one server instruction, when executed by the at least one server processor, causes the server to:

generate a first target super-resolution neural network model of a streamable media corresponding to a first target super-resolution by providing an instance of the streamable media at a first resolution as a training input to the first target super-resolution neural network model and by providing an instance of the streamable media at the first target super-resolution as a learning objective of the first target super-resolution neural network model wherein the first resolution is lower than the first target super-resolution; and

store the first target super-resolution neural network model in the at least one server memory,

wherein the at least one terminal device instruction, when executed by the at least one terminal device processor, causes the terminal device to:

stream an instance of the streamable media to the terminal device via the network;

identify a streaming resolution of the instance of the streamable media streamed to the terminal device;

based on the streaming resolution being less than the first target super-resolution, acquire from the server via the network, the first target super-resolution neural network model;

reconstruct the instance of the streamable media streamed to the terminal device into an instance of the streamable media at the first target super-resolution using the first target super-resolution neural network model; and

play the reconstructed instance of the streamable media by the terminal device,

wherein the at least one server instruction, when executed by the at least one server processor, further causes the server to:

generate a plurality of target super-resolution neural network models of the streamable media; and

store the plurality of target super-resolution neural network models in the at least one server memory, and

wherein the at least one terminal device instruction, when executed by the at least one terminal device processor, further causes the terminal device to:

send a super-resolution neural network acquisition request to the server via the network, wherein the super-resolution neural network acquisition request comprises the streaming resolution and a requested target super-resolution from among a plurality of target super-resolutions, and

receive the respective target super-resolution neural network model from among the plurality of target super-resolution neural network models corresponding to the first resolution and the requested target super-resolution.

12 . The super-resolution reconstruction system of claim 11 , wherein the first resolution is the streaming resolution.

13 . The super-resolution reconstruction system of claim 11 ,

wherein each respective target super-resolution neural network model of the plurality of target super-resolution neural network models corresponds to the first resolution and a respective target super-resolution of the plurality of target super-resolutions, and

wherein the plurality of target super-resolutions includes the first target super-resolution, the plurality of target super-resolution neural network models includes the first target super-resolution neural network model, and the first resolution is lower than each respective target super-resolution of the plurality of target super-resolutions.

14 . The super-resolution reconstruction system of claim 11 , wherein the first target super-resolution neural network model is encoded with an instance of the streamable media at the streaming resolution.

15 . The super-resolution reconstruction system of claim 11 , wherein the at least one terminal device instruction, when executed by the at least one terminal device processor, causes the terminal device to:

identify a user type; and

identify, based on the identified user type, the first target super-resolution.

16 . The super-resolution reconstruction system of claim 15 , wherein the at least one terminal device instruction, when executed by the at least one terminal device processor, further causes the terminal device to:

based on identifying the user type as an ordinary-level user, identify the first target super-resolution to be a target resolution currently selected by a user, and

based on identifying the user type as an advanced-level user, identify the first target super-resolution to be a preset advanced playing resolution.