Abstract: A deep neural network-based coding system for images determines update parameters of a deep neural network model for decoding an image. These parameters are determined by an encoder and provided to a decoder to update the model of the decoder before decoding the image. This provides structural sparsity by fine-tuning only some parameters of the neural decoder. The update is done on a set of parameters selected based on the embedding representative of the coded image so that there is no need to transmit information related to the selection of the parameters to be updated. A more generic optimizer/inference engine is also described as well as an application to sound upsampling.

Abstract: A deep neural network-based coding system for images determines update parameters of a deep neural network model for decoding an image. The parameters are determined by an encoder and provided to a decoder to update the model of the decoder before decoding the image. This provides structural sparsity by fine-tuning only some parameters of the neural decoder. The update is done either on a set of predetermined parameters so that the structural sparsity is identical for all images or on a set of parameters selected based on the image to be encoded so that the structural sparsity is image specific. A new training procedure as well as an end-to-end trainable quantization are also proposed allowing to include trained parameters in a bitstream and to update parameters in the decoder.

Abstract: Systems, methods, and instrumentalities are disclosed herein for reducing the amortization gap in end-to-end image compression and/or video compression. In examples, a video decoder may obtain an entropy model indication in video data. Based on the entropy model indication, the decoder may determine an entropy model to use for decoding a current picture. The current picture may be decoded based on the determined entropy model. In examples, the entropy model indication may indicate whether to use an updated entropy model or a prior entropy model for decoding the current picture. In examples, the entropy model indication may indicate an updated entropy model or a learned entropy model to use for decoding the current picture.

Abstract: A method or apparatus that combines properties of parallel distribution and flexibility of distribution of convolutional neural networks on heterogeneous devices uses parallelization for distributed inference of Deep Neural Networks. The method addresses parallel distribution among various numbers of devices of various capacities. A network is distributed in parallel for each configuration and is adaptive to different network capabilities.

Abstract: The proposed approach deals with efficient transmission for distributed AI with a provision to switch among multiple bandwidths. During the distributed inference at edge devices, each device needs to load part of the AI model only once, but the input/output features communicated between them can be flexibly configured depending on the available transmission bandwidth by enabling/disabling connection between nodes in the Dynamic feature size Switch (DySw). When some nodes are connected or disconnected in order to achieve the desired compression factor, other parameters of the DNN remain the same. That is, the same DNN model is used for different compression factors, and no new DNN model needs to be downloaded to adapt to the compression factor or the network bandwidth.

Abstract: An apparatus, system or method for processing a sequence of data can involve determining an availability of a computational resource; adapting, based on the availability, a neural network to set a limit on use of the computational resource by the neural network during processing of a sequence of data, wherein the one or more processors being configured to adapt the neural network to set the limit comprises the one or more processors being configured to modify an update control included in the neural network based on a windowing function; and processing at least a portion of the sequence of data by the adapted neural network using the computational resource in accordance with the limit.

Abstract: In one implementation, the AI/ML model is first split into several unitary chunks that correspond to sub-parts of the model. Then an aggregation of unitary chunks is made by considering the download time, inference time of unitary chunks, and/or device constraints. The first split corresponds to a first chunk of AI/ML layers that, once downloaded, is useable as is, and generates intermediate results based on some sensing/perception data. As soon as a new chunk arrives, it is used to generate new results based on the intermediate data of the previous chunk. Since download and inference are parallelized, a final result can be generated earlier than with the full sequential method. In addition, as soon as the inference ends on a chunk, this chunk may be removed from the device. Several AI/ML model split methods are provided to generate model subsets/chunks for different model architectures.

Abstract: In general, at least one example of an embodiment can involve determining a constraint associated with processing a sequence of data, adapting a neural network based on the constraint, wherein adapting the neural network comprises modifying, based on the constraint, a characteristic of a decision function included in the neural network; and enabling processing of at least a first portion of the sequence of data utilizing the adapted neural network and in accordance with the constraint.

Abstract: In general, at least one example of an embodiment can involve selecting a neural network from a plurality of neural networks based on an indication of resource availability and processing data using the selected neural network in accordance with the resource availability.

Abstract: A method for forwarding a connection at an interconnecting device is described. The method includes receiving from an originating device an incoming connection at an incoming port, determining by at least one hardware processor whether port translation is implemented for the incoming port, in case port translation is implemented for the incoming port, returning to the originating device by the at least one hardware processor a challenge intended to distinguish humans from computers, verifying by the at least one hardware processor that a response received in response to the challenge is a correct response to the challenge, and, in case the response is a correct response to the challenge, forwarding, by the at least one hardware processor, the connection to a translated port corresponding to the incoming port according to the port translation.

Abstract: To determine if a device is open to the Internet, the device, storing a statistically unique identifier for the device or a group of devices including the device, sends a search request comprising the statistically unique identifier to an Internet search engine, receives a response to the search request from the Internet search engine, determines from the search response whether the statistically unique identifier was found by the Internet search engine, and in case the statistically unique identifier was found found by the Internet search engine, performs an action intended to disable access to the statistically unique identifier from the Internet via the communication interface. The device can also update statistically unique identifier in case the statistically unique identifier was found in the search response.

Abstract: A method for forwarding a connection at an interconnecting device is described. The method includes receiving from an originating device an incoming connection at an incoming port, determining by at least one hardware processor whether port translation is implemented for the incoming port, in case port translation is implemented for the incoming port, returning to the originating device by the at least one hardware processor a challenge intended to distinguish humans from computers, verifying by the at least one hardware processor that a response received in response to the challenge is a correct response to the challenge, and, in case the response is is a correct response to the challenge, forwarding, by the at least one hardware processor, the connection to a translated port corresponding to the incoming port according to the port translation.

Abstract: A method and apparatus for performing story analysis are described including accepting a story, segmenting the received story into a plurality of scenes, detecting characters for each scene in the story, analyzing a relationship of each set of characters in the each scene of the story, by parsing, tagging and filtering descriptive text and words of dialog between the each set of characters to calculate a number of dialogs between each the set of characters and a number of words in each dialog between each the set of characters, determining an importance of each character in the each scene of the story, determining an interaction characterization for the each character in the each scene of the story using the importance of the each character and generating character relationship data responsive to the importance of each character in the each scene of the story and the interaction characterization for the each character in the each scene of the story.

Abstract: A method and an apparatus for generating an explanation for a recommendation based on the content of a recommended multimedia content item and a multimedia content item forming the basis for the recommendation are described. Concepts associated to the recommended multimedia content item are compared with vertices of a directed graph to associate the recommended item to a first vertex of the directed graph. The vertices of the directed graph represent concepts derived from one or more knowledge bases. Then a multimedia content item that has been previously accessed by the user is selected. Concepts associated to this previously accessed multimedia content item are compared with vertices of the directed graph to associate the previously accessed multimedia content item to a second vertex of the directed graph. Subsequently a path between the first vertex of the directed graph and the second vertex of the directed graph is determined using edges of the directed graph.

Abstract: A method and an apparatus for providing information for a multimedia content item from a catalog of multimedia content items to a user with the capability of displaying intra-content and inter-content information are described. A sequence of concepts is generated for the multimedia content item using a concept graph and metadata associated to the multimedia content item. An enhanced sequence of concepts for the multimedia content item using the sequence of concepts and an enhanced concept graph. The enhanced sequence of concepts is then displayed to a user.

Abstract: The invention concerns a method for enriching a multimedia content defined by a timeline and by a chronological text description, wherein it comprises the following steps of: identifying using natural language processing at least one feature in at least a part of a text document retrieved from a network, automatically aligning said determined part of the text document to at least a part of the chronological text description, which semantically corresponds to said determined part of the text document, so that at least the part of the text document is automatically synchronized with the timeline of the multimedia content, the chronological text description being itself synchronized with the timeline.

Abstract: The present invention relates to a method for navigating in a sequence of images. An image is displayed on a screen. A first object of the displayed image is selected at a first position according to a first input. The first object is moved to a second position according to a second input. At least one image is identified in the sequence of images where the first object is close to the second position. Playback of the sequence of images is started beginning at one of the identified images.