Abstract: According to a first aspect, there is provided a method of compressing a set of weights for a neural network, the method comprising: receiving, from a source, a plurality of weights of a neural network in a first number format, wherein a weight is represented in the first number format by N bits comprising a plurality of bits representing an exponent of the weight, a plurality of bits representing a mantissa of the weight and one bit representing a sign of the weight; and converting the plurality of weights into a second number format, wherein a weight is represented in the second number format by bw bits comprising a plurality of bits representing the exponent of the weight, a plurality of bits representing the mantissa of the weight and one bit representing a sign of the weight, wherein bw<N.

Abstract: There is provided a method for obtaining a neural network for use in data compression. The method includes initializing an encoder-side neural network at an encoding system; initializing a decoder-side neural network at a decoding system; herein the encoder-side and decoder-side neural networks have the same architecture and starting parameters; obtaining, by the encoding system, a training data item; encoding, by the encoding system, the training data item to generate a compressed training data item; transmitting, by the encoding system, the compressed training data item; receiving, by the decoding system, the compressed training data item; and training the encoder-side and decoder-side neural networks based upon the compressed training data item such that the encoder-side and decoder-side neural networks remain identical.

Abstract: Methods of processing at least one of image or video data, and a decoder are provided. The method includes successively decoding a plurality of fragments of an image or video. The decoding includes, for one or more fragment in the plurality of fragments: decoding a respective encoded residual fragment to obtain a decoded residual fragment; generating, using a machine-learning model, a predicted fragment from a respective reference fragment, where the respective reference fragment is selected from one or more reference fragments stored in a set of reference fragments; generating, based on the predicted fragment and the decoded residual fragment, a reconstructed fragment; storing the reconstructed fragment in the set of reference fragments; and updating parameters of the machine-learning model based on the respective reference fragment and the reconstructed fragment.

Abstract: A method for decoding parameters of a machine learning model is provided. The method includes that: a bitstream is decoded to determine values of samples in an image data; one or more parameter values are determined based on the values of the samples in the image data; and one or more parameters of the machine learning model are determined based upon the one or more parameter values.

Abstract: A method of processing video data, performed by a decoder, is provided. The method includes that: a bitstream including a first picture retargeted according to a first retargeting scheme and a second picture retargeted according to a second retargeting scheme is received; the first picture is decoded from the bitstream to generate a first decoded picture; the first decoded picture is adapted to the second retargeting scheme to generate a reference picture; the second picture is decoded from the bitstream to generate a second decoded picture, wherein the operation of decoding the second picture includes inter prediction using the reference picture; and an output sequence including the first and second decoded pictures is generated.

Abstract: Disclosed herein are an immersive video encoding method and an immersive video decoding method. The immersive video encoding method includes setting basic view images among input images corresponding to multiple views, generating an atlas image using the basic view images, performing encoding on the atlas image, and generating metadata about an encoded atlas image, wherein the atlas image includes a texture atlas or a depth atlas.

Abstract: An image encoding method according to the present disclosure may include classifying viewpoint images into a basic image and an additional image; performing pruning on an additional image based on the classification result, generating an atlas by packing patches obtained as a result of performing the pruning; and encoding the atlas and metadata for the atlas. In this case, a margin may be set for a patch within the atlas.

Abstract: A method of encoding video includes encoding a texture image, and encoding metadata for the texture image. A value of a chrominance sample expressed within an original dynamic range in the texture image is changed to a value within a changed dynamic range and encoded.

Abstract: A video encoding method includes classifying a plurality of view images into a basic image and additional images, performing pruning on at least one of the plurality of view images on the basis of the classification result, generating an atlas on the basis of the pruning result, and encoding the atlas and metadata for the atlas. Here, the metadata includes a first flag indicating whether depth estimation needs to be performed on a decoder side.

Abstract: Systems and methods for processing a video are disclosed herein. The method include (1) receiving an input video; (2) identifying one or more objects in the input video; (3) determining a set of feature descriptors associated with the one or more objects; (4) generating a set of feature units associated with the one or more objects.

Abstract: A transmitter system for processing a video includes: an object recognition component configured to identify one or more objects in the video and extract one or more features associated with the one or more objects; a video processing component configured to process each frame of the video by removing the one or more objects; a video encoding component configured to encode the processed video; and a transmitting component configured to transmit the encoded video and the extracted features of the one or more objects.

Abstract: A video encoding method includes classifying a plurality of view images into a basic image and additional images, performing pruning on at least one of the plurality of view images on the basis of the classification result, generating an atlas on the basis of the pruning result, and encoding the atlas and metadata for the atlas. Here, the metadata includes a first flag indicating whether depth estimation needs to be performed on a decoder side.

Abstract: A method for multiview picture data encoding comprising the steps of: performing extraction of features from multiview picture data to obtain a plurality of feature maps; performing stitching and/or transforming of the obtained plurality of feature maps to obtain at least one panoramic map of features; performing transforming of the multiview picture data to select a plurality of patches of view of the multiview picture data; encoding the at least one panoramic map of features; and encoding the plurality of patches of view.

Abstract: A method for multiview video data encoding is provided. The method includes: performing feature detection on first picture data relating to a first view to obtain a first set of features corresponding to said first view; generating a picture bitstream based on the first picture data relating to the first view; performing feature detection on second picture data relating to a second view to obtain a second set of features corresponding to said second view; performing feature matching of the first and second sets of features so as to identify an area of common characteristics; and performing prediction on second input picture data based on the area of common characteristics so as to generate a residual data bitstream.

Abstract: Disclosed herein are an immersive video decoding method and an immersive video encoding method. The immersive video decoding method includes partitioning a current picture into multiple main blocks, determining whether each main block is to be partitioned into multiple sub-blocks, and when it is determined that each main block is to be partitioned into multiple sub-blocks, determining a partitioning type of the corresponding main block. Here, the corresponding main block may be partitioned into four or two sub-blocks depending on the partitioning type. When the corresponding main block is partitioned into multiple sub-blocks, whether depth information is to be updated for each of the sub-blocks generated by partitioning the main block may be determined.

Abstract: Methods and devices for video data decoding and encoding are provided. The method for video data decoding includes: obtaining a picture bitstream; obtaining a feature bitstream indicating a residual set of features as a result of subtracting a second set of features detected in encoded picture data generated from original picture data by encoding from a first set of features detected in the original picture data; retrieving a decoded set of features from decoding the picture bitstream; and recovering the first set of features indicating the features detected in the input picture data from the decoded set of features and the residual set of features decoded from the feature bitstream.

Abstract: A visual feature processing method in an encoding device is disclosed. The visual feature processing method comprises: performing feature extraction from picture data to be encoded based on a predetermined feature extraction method to thereby obtain a set of extracted features; sorting the features in the set of extracted features based on a predetermined criterion; iteratively dividing the sorted set of extracted features in a plurality of subsets of features, said plurality of subsets of features comprising a first subset of features and at least one further subset of features, wherein the first subset of features is assigned a priority value which is higher than the priority value of the at least one further subset of features; and multiplexing the features of each subset of features for outputting for compressing, wherein the multiplexing is based on the priority value assigned to each subset of features.

Abstract: Computer-implemented methods, computer-readable media and devices for encoding video data using picture synthesis from features are provided. A computer-implemented method of encoding video data includes extracting features from a picture in a video; obtaining a predicted value of one or more regions in the picture by applying generative picture synthesis onto the features; obtaining a residual value of the one or more regions in the picture based on an original value of the one or more regions in the picture and the predicted picture; and encoding the residual value and the extracted features. Disclosed herein are also computer-implemented methods, computer-readable media and devices for decoding video data using picture synthesis from features.

Abstract: A method of encoding an immersive image according to the present disclosure comprises classifying a plurality of view images into a basic image and an additional image, generating a plurality of texture atlases based on the plurality of view images, generating a first depth atlas including depth information of view images included in a first texture atlas among the plurality of texture atlases, and generating a second depth atlas including depth information of view images included in remaining texture atlases other than the first texture atlas.

Abstract: A method for processing an immersive video includes: performing pruning for view images; generating an atlas by packing a patch that is extracted as a result of the pruning; deriving an offset for the patch that is comprised in the atlas; and correcting pixel values in the patch by using the derived offset.