Supplemental enhancement information messages for neural network based video post processing
A method, computer program, and computer system is provided for video post-processing using supplemental enhancement information (SEI) messages by determining whether a structure of a neural network and one or more parameters associated with the neural network are defined for a decoder and requesting data corresponding to the structure of the neural network and the one or more parameters based on the determination using one or more SEI messages. The requested structure and parameter data is received, and post processing is performed on video data based on the received structure and parameter data.
1. A method of video post-processing using supplemental enhancement information (SEI) messages, the method comprising:
requesting data corresponding to a structure of a neural network and data corresponding to one or more parameters based on one or more SEI messages, wherein one of the one or more parameters include an input packing type corresponding to one or more luma blocks of reconstructed luma samples and one or more chroma blocks of reconstructed chroma samples,
wherein based on the structure of the neural network and the one or more parameters being pre-defined for a decoder, the data comprises an identifier corresponding to the structure data and parameter data, and
wherein based on the structure of the neural network and the one or more parameters not being pre-defined for the decoder, the structure data and parameter data are received from a bitstream;
receiving the requested structure data and parameter data; and
performing post-processing of video data based on the received structure data and parameter data.
2. The method of claim 1 , wherein the requested data comprises an index indicating whether a predefined neural network structure is used.
3. The method of claim 1 , wherein the requested data comprises a total number of layers in the neural network, a size value associated with a convolutional kernel of the neural network, one or more special operations for each of the layers of the neural network, and a residual learning flag associated with the neural network.
4. The method of claim 1 , wherein the requested data comprises a flag indicating whether luma data and chroma data associated with the video data use different neural networks.
5. The method of claim 1 , wherein the requested data comprises an identifier indicating a total number of neural network parameter sets.
6. The method of claim 1 , wherein the requested data comprises one or more bit-depth precision values and bit-depth precision description identifiers for the parameter data.
7. The method of claim 1 , wherein the requested data comprises a determination of whether the transmitted parameters are compressed and quantized.
8. The method of claim 1 , wherein the requested data comprises an identifier indicating whether side information is used, wherein the side information is used to calculate a weight mask in the filter process.
9. The method of claim 1 , wherein the requested data comprises a block-level supporting control size value, wherein the block-level support control size value is used to toggle use of one or more neural network tools and parameter set switching at the block level.
10. A computer system for video post-processing using supplemental enhancement information (SEI) messages, the computer system comprising:
one or more computer-readable non-transitory storage media configured to store computer program code; and
one or more computer processors configured to access said computer program code and operate as instructed by said computer program code, said computer program code including:
requesting code configured to cause the one or more computer processors to request data corresponding to a structure of a neural network and data corresponding to one or more parameters based on one or more SEI messages, wherein one of the one or more parameters include an input packing type corresponding to one or more luma blocks of reconstructed luma samples and one or more chroma blocks of reconstructed chroma samples,
wherein based on the structure of the neural network and the one or more parameters being pre-defined for a decoder, the data comprises an identifier corresponding to the structure data and parameter data, and
wherein based on the structure of the neural network and the one or more parameters not being pre-defined for the decoder, the structure data and parameter data are received from a bitstream;
receiving code configured to cause the one or more computer processors to receive the requested structure data and parameter data; and
performing code configured to cause the one or more computer processors to perform post-processing of video data based on the received structure data and parameter data.
11. The computer system of claim 10 , wherein the requested data comprises an index indicating whether a predefined neural network structure is used.
12. The computer system of claim 10 , wherein the requested data comprises a total number of layers in the neural network, a size value associated with a convolutional kernel of the neural network, one or more special operations for each of the layers of the neural network, and a residual learning flag associated with the neural network.
13. The computer system of claim 10 , wherein the requested data comprises a flag indicating whether luma data and chroma data associated with the video data use different neural networks.
14. The computer system of claim 10 , wherein the requested data comprises an identifier indicating a total number of neural network parameter sets.
15. The computer system of claim 10 , wherein the requested data comprises one or more bit-depth precision values and bit-depth precision description identifiers for the parameter data.
16. The computer system of claim 10 , wherein the requested data comprises an identifier indicating whether side information is used, wherein the side information is used to calculate a weight mask in the filter process.
17. The computer system of claim 10 , wherein the requested data comprises a block-level supporting control size value, wherein the block-level support control size value is used to toggle use of one or more neural network tools and parameter set switching at the block level.
18. A non-transitory computer readable medium having stored thereon a computer program for video post-processing using supplemental enhancement information (SEI) messages, the computer program configured to cause one or more computer processors to:
request data corresponding to a structure of a neural network and data corresponding to one or more parameters based on one or more SEI messages, wherein one of the one or more parameters include an input packing type corresponding to one or more luma blocks of reconstructed luma samples and one or more chroma blocks of reconstructed chroma samples,
wherein based on the structure of the neural network and the one or more parameters being pre-defined for a decoder, the data comprises an identifier corresponding to the structure data and parameter data, and
wherein based on the structure of the neural network and the one or more parameters not being pre-defined for the decoder, the structure data and parameter data are received from a bitstream;
receive the requested structure data and parameter data; and
perform post-processing on video data based on the received structure data and parameter data.