Content-adaptive online training with scaling factors and/or offsets in neural image compression

1. A method for video decoding in a video decoder, the method comprising:

decoding neural network update information in a coded bitstream for at least one neural network in the video decoder, the at least one neural network being configured with a set of pretrained parameters corresponding to a first set of training images, the neural network update information indicating a first modification parameter indicating a scaling factor or offset applicable to the pretrained parameters based on a second set of training images different from the first set of training images;

updating the set of pretrained parameters in the at least one neural network in the video decoder by applying the scaling factor or offset of the first modification parameter to the pretrained parameters; and

decoding an encoded image based on the updated at least one neural network.

2. The method of claim 1 , wherein

the first modification parameter is the scaling factor, and

the updating includes replacing the set of pretrained parameters with a set of replacement parameters, respectively, the set of replacement parameters being the set of pretrained parameters multiplied with the scaling factor.

3. The method of claim 1 , wherein

the first modification parameter is the offset, and

the updating includes adding the offset to the set of pretrained parameters in the at least one neural network.

4. The method of claim 2 , wherein

the neural network update information further indicates the offset to modify the set of pretrained parameters, and

the updating includes adding the offset to the set of replacement parameters in the at least one neural network.

5. The method of claim 1 , wherein

the coded bitstream further indicates one or more encoded bits directed to a context model for decoding the encoded image,

the at least one neural network includes a main decoder network, a context model neural network, an entropy parameter neural network, and a hyper decoder network in the video decoder,

the method further includes:

decoding the one or more encoded bits using the hyper decoder network, and

determining the context model using the context model neural network and the entropy parameter neural network based on the decoded bits and a quantized latent of the encoded image that is available to the context model neural network, and

the decoding the encoded image includes decoding the encoded image with the main decoder network and the context model.

6. The method of claim 1 , wherein the at least one neural network is one or more of a main decoder network, a context model neural network, an entropy parameter neural network, or a hyper decoder network in the video decoder.

7. The method of claim 1 , wherein the set of pretrained parameters includes biases in a convolutional layer of one of the at least one neural network.

8. The method of claim 1 , wherein

the set of pretrained parameters includes weights in a convolutional kernel of a convolutional layer of one of the at least one neural network.

9. The method of claim 1 , wherein

the neural network update information indicates a second modification parameter for the at least one neural network configured with a second set of pretrained parameters, and

the updating includes updating the second set of pretrained parameters in the at least one neural network based on the second modification parameter.

10. The method of claim 9 , wherein

the set of pretrained parameters includes weights in a convolutional kernel of a convolutional layer of one of the at least one neural network, and

the second set of pretrained parameters includes biases of the convolutional layer of the one of the at least one neural network.

11. The method of claim 1 , further comprising:

decoding another encoded image in the coded bitstream based on the updated at least one neural network.

12. An apparatus for video decoding, comprising:

processing circuitry configured to:

decode neural network update information in a coded bitstream for at least one neural network in the apparatus for video decoding, the at least one neural network being configured with a set of pretrained parameters corresponding to a first set of training images, the neural network update information indicating a first modification parameter indicating a scaling factor or offset applicable to the pretrained parameters based on a second set of training images different from the first set of training images;

update the set of pretrained parameters in the at least one neural network in the apparatus for video decoding by applying the scaling factor or offset of the first modification parameter to the pretrained parameters; and

decode an encoded image based on the updated at least one neural network.

13. The apparatus of claim 12 , wherein

the first modification parameter is the scaling factor, and

the processing circuitry is configured to replace the set of pretrained parameters with a set of replacement parameters, respectively, the set of replacement parameters being the set of pretrained parameters multiplied with the scaling factor.

14. The apparatus of claim 12 , wherein

the first modification parameter is the offset, and

the processing circuitry is configured to add the offset to the set of pretrained parameters in the at least one neural network.

15. The apparatus of claim 13 , wherein

the neural network update information further indicates the offset to modify the set of pretrained parameters, and

the processing circuitry is configured to add the offset to the set of replacement parameters in the at least one neural network.

16. The apparatus of claim 12 , wherein the set of pretrained parameters includes biases in a convolutional layer of one of the at least one neural network.

17. The apparatus of claim 12 , wherein

the set of pretrained parameters includes weights in a convolutional kernel of a convolutional layer of one of the at least one neural network.

18. The apparatus of claim 12 , wherein

the neural network update information indicates a second modification parameter for the at least one neural network configured with a second set of pretrained parameters, and

the processing circuitry is configured to update the second set of pretrained parameters in the at least one neural network based on the second modification parameter.

19. The apparatus of claim 18 , wherein

the set of pretrained parameters includes weights in a convolutional kernel of a convolutional layer of one of the at least one neural network, and

the second set of pretrained parameters includes biases of the convolutional layer of the one of the at least one neural network.

20. A non-transitory computer-readable storage medium storing a program executable by at least one processor to perform:

decoding neural network update information in a coded bitstream for at least one neural network in a video decoder, the at least one neural network being configured with a set of pretrained parameters corresponding to a first set of training images, the neural network update information indicating a first modification parameter indicating a scaling factor or offset applicable to the pretrained parameters based on a second set of training images different from the first set of training images;

updating the set of pretrained parameters in the at least one neural network in the video decoder by applying the scaling factor or offset of the first modification parameter to the pretrained parameters; and

decoding an encoded image based on the updated at least one neural network.