公开(公告)号：US20220405979A1

公开(公告)日：2022-12-22

申请号：US17826806

申请日：2022-05-27

申请人： Tencent America LLC

发明人： Ding DING ,  Wei JIANG ,  Wei WANG ,  Shan LIU

IPC分类号： G06T9/00 ,  H04N19/86 ,  H04N19/44 ,  G06N3/08

摘要： Aspects of the disclosure provide a method, an apparatus, and non-transitory computer-readable storage medium for video decoding. The apparatus includes processing circuitry that reconstructs blocks of an image that is to be reconstructed from a coded video bitstream. The processing circuitry decodes first deblocking information in the coded video bitstream including a first deblocking parameter of a deep neural network (DNN) in a video decoder. The first deblocking parameter of the DNN is an updated parameter that has been previously determined by a content adaptive training process. The processing circuitry determines the DNN for a first boundary region comprising a subset of samples in the reconstructed blocks based on the first deblocking parameter included in the first deblocking information. The processing circuitry deblocks the first boundary region comprising the subset of samples in the reconstructed blocks based on the determined DNN corresponding to the first deblocking parameter.

公开(公告)号：US20220335656A1

公开(公告)日：2022-10-20

申请号：US17703527

申请日：2022-03-24

申请人： TENCENT AMERICA LLC

发明人： Wei JIANG ,  Wei WANG ,  Xiaozhong XU ,  Shan LIU

IPC分类号： G06T9/00 ,  G06T3/40 ,  G06N3/08

摘要： A method and apparatus for adaptive neural image compression with smooth quality control by meta-learning includes receiving an input image and a target quality control parameter; generating quality-adaptive weight parameters of an encoding neural network using shared encoding parameters and adaptive encoding parameters; and encoding the input image, based on the target quality control parameter, using the encoding neural network with the quality-adaptive weight parameters, to generate a compressed representation.

公开(公告)号：US20220335655A1

公开(公告)日：2022-10-20

申请号：US17702177

申请日：2022-03-23

申请人： TENCENT AMERICA LLC

发明人： Wei JIANG ,  Wei WANG ,  Xiaozhong XU ,  Shan LIU

IPC分类号： G06T9/00

摘要： The present disclosure includes a method, apparatus, and non-transitory computer-readable medium for adaptive neural image compression by meta-learning. The method may include generating a substitute input image and a substitute target quality control parameter using an original input image and a target quality control parameter, wherein the substitute input image is a modified version of the original input image and the substitute target quality control parameter is a modified version of the target quality control parameter. The method may further include encoding the substitute input image, based on the substitute input image and the substitute target quality control parameter, using an encoding neural network, to generate a compressed representation of the substitute input image.

公开(公告)号：US20220224901A1

公开(公告)日：2022-07-14

申请号：US17486239

申请日：2021-09-27

申请人： TENCENT AMERICA, LLC

发明人： Wei JIANG ,  Wei WANG ,  Sheng LIN ,  Shan LIU

IPC分类号： H04N19/124 ,  H04N19/117 ,  G06N3/04

摘要： Video processing with a multi-quality loop filter using a multi-task neural network is performed by at least one processor and includes generating model IDs, based on quantization parameters in an input, selecting a first set of masks, each mask in the first set of masks corresponding to one of the generated model IDs, performing convolution of first weights of a first set of neural network layers and the selected first set of masks to obtain first masked weights, and selecting a second set of neural network layers and second weights, based on the quantization parameters, generating a quantization parameter value, based on the quantization parameters, and computing an inference output, based on the first masked weights and the second weights, using the generated quantization parameter value.

公开(公告)号：US20220215265A1

公开(公告)日：2022-07-07

申请号：US17478089

申请日：2021-09-17

申请人： TENCENT AMERICA LLC

发明人： Wei JIANG ,  Wei WANG ,  Sheng LIN ,  Shan LIU

IPC分类号： G06N3/08 ,  G06N3/04

摘要： End-to-end task oriented latent compression using deep reinforcement learning (DRL) is performed by at least one processor and includes generating latent representations of an input image using a first neural network, wherein the latent representations is a sequence of latent signals, encoding the latent signals using a second neural network, generating a set of quantization keys based on a set of previous quantization states, wherein each quantization key in the set of quantization keys and each previous quantization state in the set of previous quantization states correspond to each of the latent signals using a third neural network, generating a set of dequantized numbers representing dequantized representations of the encoded latent signals based on the set of quantization keys using a fourth neural network, generating a reconstructed output based on the set of dequantized numbers, and performing a target task based on the reconstructed output using a fifth neural network.

公开(公告)号：US20220174281A1

公开(公告)日：2022-06-02

申请号：US17488438

申请日：2021-09-29

申请人： TENCENT AMERICA LLC

发明人： Wei JIANG ,  Wei WANG ,  Shan LIU

IPC分类号： H04N19/124 ,  G06N3/04 ,  G06N3/08

摘要： There is included a method and apparatus comprising computer code configured to cause a processor or processors to perform obtaining an input stream of video data, computing a key based on a floating number in the input stream, predicting a current dependent quantization (DQ) state based on a state predictor and a number of previous keys and a number of previous DQ states, reconstructing the floating number based on the key and the current DQ state, and coding the video based on the reconstructed floating number.

公开(公告)号：US20220116610A1

公开(公告)日：2022-04-14

申请号：US17559676

申请日：2021-12-22

申请人： TENCENT AMERICA LLC

发明人： Wei WANG ,  Wei JIANG ,  Shan LIU

IPC分类号： H04N19/124 ,  H04N19/13 ,  H04N19/119 ,  H04N19/192 ,  H04N19/46 ,  G06N3/08 ,  H04N19/597 ,  H04N19/96 ,  H04N19/30 ,  H04N19/176 ,  H04N19/147

摘要： A method of quantization, adaptive block partitioning and codebook coding for neural network model compression, is performed by at least one processor and includes determining a saturated maximum value of a multi-dimensional tensor in a layer of a neural network, and a bit depth corresponding to the saturated maximum value, and clipping weight coefficients in the multi-dimensional tensor to be within a range of the saturated maximum value. The method further includes quantizing the clipped weight coefficients, based on the bit depth, and transmitting, to a decoder, a layer header including the bit depth.

公开(公告)号：US20210326710A1

公开(公告)日：2021-10-21

申请号：US17225486

申请日：2021-04-08

申请人： TENCENT AMERICA LLC

发明人： Wei WANG ,  Wei JIANG ,  Shan LIU

IPC分类号： G06N3/08 ,  H03M7/30

摘要： Methods and apparatuses of neural network model compression/decompression are described. In some examples, an apparatus of neural network model decompression includes receiving circuitry and processing circuitry. The processing circuitry can be configured to receive a dependent quantization enabling flag from a bitstream of a compressed representation of a neural network. The dependent quantization enabling flag can indicate whether a dependent quantization method is applied to model parameters of the neural network. The model parameters of the neural network can be reconstructed based on the dependent quantization method in response to the dependent quantization enabling flag indicating the dependent quantization method is used for encoding the model parameters of the neural network.

公开(公告)号：US20210159912A1

公开(公告)日：2021-05-27

申请号：US17081642

申请日：2020-10-27

申请人： Tencent America LLC

发明人： Wei WANG ,  Wei JIANG ,  Shan LIU

IPC分类号： H03M7/30 ,  G06N3/02

摘要： Aspects of the disclosure provide methods and apparatuses for neural network model compression/decompression. In some examples, an apparatus for neural network model decompression includes receiving circuitry and processing circuitry. The processing circuitry decodes, from a bitstream corresponding to a representation of a neural network, at least a syntax element to be applied to multiple blocks in the neural network. Then, the processing circuitry reconstructs, from the bitstream, weight coefficients in the blocks based on the syntax element.

公开(公告)号：US20210090217A1

公开(公告)日：2021-03-25

申请号：US17023055

申请日：2020-09-16

申请人： Tencent America LLC

发明人： Wei JIANG ,  Wei WANG ,  Shan LIU

IPC分类号： G06T3/40 ,  G06N3/04 ,  G06K9/62

摘要： A video super resolution (SR) method based on video coding for machine (VCM) is provided for an electronic device. The method includes obtaining a lower resolution (LR) video; generating feature representations of the LR video based on a deep neural network (DNN); encoding the feature representations of the LR video, based on a VCM standard, and the LR video to form encoded feature representations of the LR video and an encoded LR video, wherein the feature representations of the LR video contain both space and temporal information on the LR video for creating a high resolution (HR) video corresponding to the LR video; and storing and transferring the encoded feature representations of the LR video and an encoded LR video for decoding.