A Deep Learning Method for Detecting Inappropriate Contents in A Visual Media


    Prof. Dr. Hezerul Abdul Karim
  • Description of Invention

    The present invention relates to a method for detecting inappropriate contents in a visual media. More particularly, the method is developed to analyze inappropriate contents, such as nudity and pornography, in a visual media that include videos and images. The method takes advantages on development of artificial intelligence and machine learning model for obtaining accurate result in detecting and filtering inappropriate contents of visual media with minimum human intervention. The present invention implements a convolutional neural network, CNN model and a human detection model, EfficientDet-D7, which integrates a CNN model with weighted bi-directional feature pyramid network, biFPN model configured to detect even the smallest scale of inappropriate contents and classify the visual media under inappropriate category.

  • Intellectual Property (IP) Status

    • Patent Filed
    • TRL Status: 4