-
公开(公告)号:US11477485B2
公开(公告)日:2022-10-18
申请号:US17510519
申请日:2021-10-26
Inventor: Takahiro Nishi , Tadamasa Toma , Kiyofumi Abe , Ryuichi Kanoh , Luca Rigazio , Alec Hodgkinson
IPC: H04N19/61 , H04N19/503
Abstract: The encoder includes processing circuitry, and memory. Using the memory, the processing circuitry: generates a predicted image of an input image that is a current image to be encoded, based on generated data output from a generator network in response to a reference image being input to the generator network, the generator network being a neural network; calculates a prediction error by subtracting the predicted image from the input image; and generates an encoded image by at least transforming the prediction error.
-
公开(公告)号:US11166032B2
公开(公告)日:2021-11-02
申请号:US17080141
申请日:2020-10-26
Inventor: Alec Hodgkinson , Luca Rigazio , Takahiro Nishi , Kiyofumi Abe , Ryuichi Kanoh , Tadamasa Toma
IPC: H04N19/184 , H04N19/46 , H04N19/85 , H04N19/136 , H04N19/124 , G06N3/08
Abstract: An encoder includes circuitry and memory. Using the memory, the circuitry: encodes an original image and decodes the original image encoded, to generate a first bitstream and a local decoded image; encodes supplemental information and decodes the encoded supplemental information, to generate a second bitstream and local decoded supplemental information; inputs data based on the local decoded image and the local decoded supplemental information to a post processing network which is a neural network, to cause a reconstructed image to be output from the post processing network, the reconstructed image corresponding to the original image and being to be used to encode a following original image which follows the original image; and concatenates the first bitstream and the second bitstream to generate a concatenated bitstream.
-
公开(公告)号:US11995150B2
公开(公告)日:2024-05-28
申请号:US17234127
申请日:2021-04-19
Inventor: Denis Gudovskiy , Alec Hodgkinson , Takuya Yamaguchi , Sotaro Tsukizawa
IPC: G06F18/214 , G06N5/04 , G06N20/00
CPC classification number: G06F18/214 , G06N5/04 , G06N20/00
Abstract: An information processing method implemented by a computer includes: obtaining a piece of first data, and a piece of second data not included in a training dataset for training an inferencer; calculating, using a piece of first relevant data obtained by inputting the first data to the inferencer trained by machine learning using the training dataset, a first contribution representing contributions of portions constituting the first data to a piece of first output data output by inputting the first data to the inferencer; calculating, using a piece of second relevant data obtained by inputting the second data to the inferencer, a second contribution representing contributions of portions constituting the second data to a piece of second output data output by inputting the second data to the inferencer; and determining whether to add the second data to the training dataset, according to the similarity between the first and second contributions.
-
公开(公告)号:US11190804B2
公开(公告)日:2021-11-30
申请号:US16664084
申请日:2019-10-25
Inventor: Takahiro Nishi , Tadamasa Toma , Kiyofumi Abe , Ryuichi Kanoh , Luca Rigazio , Alec Hodgkinson
IPC: H04N19/61 , H04N19/503
Abstract: The encoder includes processing circuitry, and memory. Using the memory, the processing circuitry: generates a predicted image of an input image that is a current image to be encoded, based on generated data output from a generator network in response to a reference image being input to the generator network, the generator network being a neural network; calculates a prediction error by subtracting the predicted image from the input image; and generates an encoded image by at least transforming the prediction error.
-
公开(公告)号:US11057646B2
公开(公告)日:2021-07-06
申请号:US16866954
申请日:2020-05-05
Inventor: Alec Hodgkinson , Luca Rigazio , Tadamasa Toma , Takahiro Nishi , Kiyofumi Abe , Ryuichi Kanoh
Abstract: An image processor includes memory and circuitry. The circuitry performs processing of approximating a decompressed image to an original image by using a neural network model trained to approximate the decompressed image to the original image. The decompressed image is obtained as a result of compression of the original image and decompression of the compressed image. The neural network model includes one or more convolutional blocks, and includes one or more residual blocks. Each of the one or more convolutional blocks is a processing block including a convolutional layer. Each of the one or more residual blocks includes a convolutional group including at least one of the one or more convolutional blocks, inputs data which is input to the residual block to the convolutional group included in the residual block, and adds the data input to the residual block to data to be output from the convolutional group.
-
-
-
-