18054274. VIDEO PROCESSING USING DELTA DISTILLATION simplified abstract (QUALCOMM Incorporated)

From WikiPatents
Jump to navigation Jump to search

VIDEO PROCESSING USING DELTA DISTILLATION

Organization Name

QUALCOMM Incorporated

Inventor(s)

Amirhossein Habibian of Amsterdam (NL)

Davide Abati of Amsterdam (NL)

Haitam Ben Yahia of Diemen (NL)

VIDEO PROCESSING USING DELTA DISTILLATION - A simplified explanation of the abstract

This abstract first appeared for US patent application 18054274 titled 'VIDEO PROCESSING USING DELTA DISTILLATION

Simplified Explanation

The patent application describes a method for processing video content using an artificial neural network. Here are the key points:

  • The method involves receiving a video data stream with multiple frames.
  • First features are extracted from the first frame using a teacher neural network.
  • The difference between the first frame and the second frame is determined.
  • Second features are extracted from the difference between the frames using a student neural network.
  • A feature map for the second frame is generated by combining the first and second features.
  • An inference is generated for the second frame based on the generated feature map.

Potential applications of this technology:

  • Video compression: The method can be used to efficiently encode video content by extracting relevant features and generating inferences for subsequent frames.
  • Video analysis: By extracting features and generating inferences, the method can aid in tasks such as object detection, tracking, and recognition in video surveillance or autonomous driving systems.
  • Video enhancement: The technique can be used to improve the quality of video content by generating inferences based on extracted features.

Problems solved by this technology:

  • Efficient processing: By using a teacher-student neural network approach, the method optimizes the extraction of features and generation of inferences, leading to more efficient video processing.
  • Information extraction: The method allows for the extraction of relevant features from video frames, enabling better analysis and understanding of the content.
  • Resource optimization: By generating inferences based on feature maps, the method reduces the computational resources required for video processing tasks.

Benefits of this technology:

  • Improved video processing efficiency: The method optimizes the extraction of features and generation of inferences, leading to faster and more efficient video processing.
  • Enhanced video analysis capabilities: By extracting relevant features and generating inferences, the method improves the accuracy and effectiveness of video analysis tasks.
  • Resource savings: The technique reduces the computational resources required for video processing, resulting in potential cost savings and improved performance.


Original Abstract Submitted

Certain aspects of the present disclosure provide techniques and apparatus for processing video content using an artificial neural network. An example method generally includes receiving a video data stream including at least a first frame and a second frame. First features are extracted from the first frame using a teacher neural network. A difference between the first frame and the second frame is determined. Second features are extracted from at least the difference between the first frame and the second frame using a student neural network. A feature map for the second frame is generated based a summation of the first features and the second features. An inference is generated for at least the second frame of the video data stream based on the generated feature map for the second feature.