Abstract: Embodiments herein provide a system and method for automatically generating a sign language video from an input speech using the machine learning model. The method includes (i) extracting a plurality of spectrograms of an input speech by (a) encoding, using an encoder, a time domain series of the input speech to a frequency domain series, and (b) decoding, using a decoder, a plurality of tokens for time steps of the frequency domain series, (ii) generating a plurality of pose sequences for a current time step of the plurality of spectrograms using a first machine learning model, and (iii) automatically generating, using a discriminator of a second machine learning model, a sign language video for the input speech using the plurality of pose sequences and the plurality of spectrograms when the plurality of pose sequences are matched with corresponding the plurality of spectrograms that are extracted.

Abstract: Embodiments herein provide a system and a method for automatically generating at least one synthetic talking head video using a machine learning model. The method includes (i) extracting features from each frame of a video that is extracted from data sources, (ii) analyzing, using a face-detection model, the video to determine a driving face video if a number of identities, and faces of speakers are equal to one in all frames of the video, (iii) generating, using a text to speech model, synthetic speech utterances by automatically selecting a vocabulary of words and sentences from the data sources, (iv) modifying lip movements that are originally present in the driving face video corresponding to the synthetic speech utterances, and (v) generating, using machine learning model, synthetic talking head video based on the lip movements that are modified corresponding to the synthetic speech utterances.

Abstract: A processor-implemented method for generating a lip-sync for a face to a target speech of a live session to a speech in one or more languages in-sync with improved visual quality using a machine learning model and a pre-trained lip-sync model is provided. The method includes (i) determining a visual representation of the face and an audio representation, the visual representation includes crops of the face; (ii) modifying the crops of the face to obtain masked crops; (iii) obtaining a reference frame from the visual representation at a second timestamp; (iv) combining the masked crops at the first timestamp with the reference to obtain lower half crops; (v) training the machine learning model by providing historical lower half crops and historical audio representations as training data; (vi) generating lip-synced frames for the face to the target speech, and (vii) generating an in-sync lip-synced frames by the pre-trained lip-sync model.

Abstract: A system for making available at an end-user a media file, from a media provider comprising a media file patch related to at least one object, the system comprising: an encoding module at the media provider configured for determining at least one representation which resembles the media file patch, by comparing the media file patch with representations of said at least one object, and for including at least one identification corresponding with said representation in a skeleton file; a storage medium storing a dictionary including the representations of the at least one object at of the end-user and an intermediate node between the media provider and/or the end-user; a decoding module configured for decoding the skeleton file using the identification for looking up the corresponding representation in the dictionary of the storage medium and for rendering the media file patch based on the looked-up corresponding representation.

Abstract: Method for capturing an environment with objects, using a 3D camera, wherein the images of the cameras captured at different moments in time are used to generate 3D models, and wherein accuracy values are assigned to segments of the models allowing efficient refining of the models using the accuracy values.

Abstract: Method for capturing an environment with objects, using a 3D camera, wherein the images of the cameras captured at different moments in time are used to generate 3D models, and wherein accuracy values are assigned to segments of the models allowing efficient refining of the models using the accuracy values.

Abstract: A Method for retrieving at least one image from a database (DB) of images based on at least two input images (image i, image j), comprises the steps of determining (120) first low level feature correspondences (LC1) between said at least two input images (image i, image j), searching (200) within said database (DB) for at least two sets (Mi, Mj) of images respectively matching said at least to images (image i, image j), determining (120) second low level features correspondences (LC2) between respective images from said at least two sets of images (Mi, Mj), determining (130) a first set of relationships (RLC1) between entities of said at least two input images (image i, image j) based on said first low level feature correspondences (LC1), determining (130) a second set of relationships (RLC2) between respective entities of said respective images from said at least two sets of images (Mi, Mj) based on said second low level feature correspondences (LC2), identifying (300) matching relationships between said fir

Abstract: A system for making available at an end-user a media file, from a media provider comprising a media file patch related to at least one object, the system comprising: an encoding module at the media provider configured for determining at least one representation which resembles the media file patch, by comparing the media file patch with representations of said at least one object, and for including at least one identification corresponding with said representation in a skeleton file; a storage medium storing a dictionary including the representations of the at least one object at of the end-user and an intermediate node between the media provider and/or the end-user; a decoding module configured for decoding the skeleton file using the identification for looking up the corresponding representation in the dictionary of the storage medium and for rendering the media file patch based on the looked-up corresponding representation.

Abstract: The present invention relates to a media content ordering system and to a method for ordering media content. According to the invention, media content items are ordered in two different spaces, i.e. metadata space and feature space. This allows a user to select and retrieve desired content more easily. Media content that is clustered in either space represents similar media content. Suggestions can be made to the user taking into account the preferences of the user with respect to features and metadata particulars. By minimizing the difference in order in both spaces, it is ensured that suggestions to a user are close both in feature space and metadata space.

Abstract: The invention relates to a system and method for tag based access to auxiliary information during a video- and/or audio-conference. The system comprises a mapping between tags and associated portions of auxiliary information. The method may comprise transmitting video data and/or or audio data from the video- and/or audio-conferencing system to participants of the video-conference. During the transmission tags may be extracted from the video data and/or audio data. As soon as a request for auxiliary information is received from a participant, the method may comprise selecting at least one of the tags extracted from the transmitted video data and/or audio data, retrieving at least one auxiliary information portion associated with the selected at least one tag, and transmitting the at least one retrieved auxiliary information portion to the participant that has requested the auxiliary information.

Abstract: Embodiments relates to a method for rotating a multidimensional space and a device for rotating a multidimensional space. The method may comprise:—obtaining (S1) 2D coordinates of a first point (p1) and a second point (p2) of a 2D visualization space input by a user,—determining (S2) a first D-dimensional vector (p1) and a second D-dimensional vector (p2) by projecting said first point (p1) and said second point (p2) into a D-dimensional space, wherein D is equal to or greater than 4,—determining (S3) a rotation matrix (R) in function of said first D-dimensional vector (p1) and said second D-dimensional vector (p2). The D-dimensional space has a first dimension, a second dimension and D-2 other dimensions, and the 2D visualization space has D-2 predefined regions associated with respective dimensions among said D-2 other dimensions.

Abstract: A process for generating a high-resolution video stream, the process comprising: receiving (310) a low-resolution video stream; receiving (320) at least one high-resolution video stream; selecting (331) first image patches from said at least one high-resolution video stream; generating (332) respective first low-resolution counterparts of said first high-resolution image patches; storing (333) said first high-resolution image patches indexed by said first low-resolution counterparts in a first data storage; and improving (350) said low-resolution video stream by substituting (351) portions of said low-resolution video stream that are similar to one or more of said first low-resolution counterparts with first high-resolution patches obtained from said first data storage in accordance with said indexing; wherein said low-resolution video stream and said at least one high-resolution video stream are substantially synchronized video streams.

Abstract: A Method for retrieving at least one image from a database (DB) of images based on at least two input images (image i, image j), comprises the steps of determining (120) first low level feature correspondences (LC1) between said at least two input images (image i, image j), searching (200) within said database (DB) for at least two sets (Mi, Mj) of images respectively matching said at least to images (image i, image j), determining (120) second low level features correspondences (LC2) between respective images from said at least two sets of images (Mi, Mj), determining (130) a first set of relationships (RLC1) between entities of said at least two input images (image i, image j) based on said first low level feature correspondences (LC1), determining (130) a second set of relationships (RLC2) between respective entities of said respective images from said at least two sets of images (Mi, Mj) based on said second low level feature correspondences (LC2), identifying (300) matching relationships between said fir

Abstract: A method for adapting video data (video room1) recorded by a camera on a location (room1) during a video conference such as to hide the presence of a participant at said location of said video conference, comprises a step of registering a predefined gesture possibly to be performed by any participant of said video conference at said location (room1), a step of detecting said gesture, and upon detection thereof, identifying the at least one participant having performed said gesture at said location (room1), adapting said video data such as to eliminate data relating to said at least one participant having performed said gesture from said video data, thereby generating adapted video data (videoa room1), for being transmitted to other participants of said video conference on other locations (room2, . . . , roomn).