Video profile creation system
AssigneeMicrosoft Technology Licensing, LLC
InventorsYurong Jiang, Federico Guillermo Hlawaczek, Karthiek Chandrasekaran
Life of this patent
Estimating the term…
Abstract
Machine learning based method and system for video profile creation are described. The technical problem of permitting users to create professional quality video profile without requiring the expertise in using specialized video creation tools is addressed by a video profile creation system powered by machine learning methodologies. In one embodiment, the video profile creation system provides a guided video recording process, and also provides a video processing pipeline, where a recorded video is transformed into a professional-looking video profile.
Claims20
A computer implemented method comprising:
- maintaining a plurality of user profiles of respective users in an online connection network system, the plurality of user profiles comprising respective profile images representing the respective users;
- accessing training data, the training data generated using the respective profile images, each image in the training data including a first set of pixels having a person pixel identifier indicating that a pixel is associated with a depiction of a person and a second set of pixels having a background pixel identifier indicating that a pixel is associated with a depiction of a background;
- training a machine learning model using the training data, the machine learning model takes a plurality of pixels as input and assigns, to each pixel in the plurality of pixels, the person pixel identifier or the background pixel identifier;
- executing the machine learning model with respect to a plurality of frames in a recorded video to generate, for frames in the plurality of frames, respective binary masks, a binary mask for a frame from the plurality of frames indicating pixels assigned the person pixel identifier and pixels assigned the background pixel identifier;
- generating a profile video by using the respective binary masks to replace, in the plurality of frames, pixels assigned the background pixel identifier with pixels of a previously stored background image;
- and
- embedding the profile video in a profile web page of the user in the online connection network system.
A system comprising:
- one or more processors;
- and
- a non-transitory computer readable storage medium comprising instructions that when executed by the one or processors cause the one or more processors to perform operations comprising:
- maintaining a plurality of user profiles of respective users in an online connection network system, the plurality of user profiles comprising respective profile images representing the respective users;
- accessing training data, the training data generated using the respective profile images, each image in the training data including a first set of pixels having a person pixel identifier indicating that a pixel is associated with a depiction of a person and a second set of pixels having a background pixel identifier indicating that a pixel is associated with a depiction of a background;
- training a machine learning model using the training data, the machine learning model takes a plurality of pixels as input and assigns, to each pixel in the plurality of pixels, the person pixel identifier or the background pixel identifier;
- executing the machine learning model with respect to a plurality of frames in a recorded video to generate, for frames in the plurality of frames, respective binary masks, a binary mask for a frame from the plurality of frames indicating pixels assigned the person pixel identifier and pixels assigned the background pixel identifier;
- generating a profile video by using the respective binary masks to replace, in the plurality of frames, pixels assigned the background pixel identifier with pixels of a previously stored background image;
- and
- embedding the profile video in a profile web page of the user in the online connection network system.
A machine-readable non-transitory storage medium having instruction data executable by a machine to cause the machine to perform operations comprising:
- a non-transitory computer readable storage medium comprising instructions that when executed by the one or processors cause the one or more processors to perform operations comprising:
- maintaining a plurality of user profiles of respective users in an online connection network system, the plurality of user profiles comprising respective profile images representing the respective users;
- accessing training data, the training data generated using the respective profile images, each image in the training data including a first set of pixels having a person label indicating a person represented in the image and a second set of pixels having a background label indicating a background represented in the image;
- training a machine learning model using the training data, the machine learning model takes a plurality of pixels as input and assigns, to each pixel in the plurality of pixels, the person label or the background label;
- executing the machine learning model with respect to a plurality of frames in a recorded video to generate a binary mask indicating pixels that represent a user in the plurality of frames and pixels that represent background in the plurality of frames;
- generating a profile video by using the binary mask to replace pixels that represent background in the plurality of frames with a previously stored background image;
- and
- embedding the profile video in a profile web page of the user in the online connection network system.
Dependencies are inferred from the claim text.