18244016. VIRTUAL OBJECT MACHINE LEARNING simplified abstract (Snap Inc.)

VIRTUAL OBJECT MACHINE LEARNING

Organization Name

Snap Inc.

Inventor(s)

Xuehan Xiong of Los Angeles CA (US)

Zehao Xue of Los Angeles CA (US)

VIRTUAL OBJECT MACHINE LEARNING - A simplified explanation of the abstract

This abstract first appeared for US patent application 18244016 titled 'VIRTUAL OBJECT MACHINE LEARNING

Simplified Explanation

Abstract: A machine learning scheme is described that can be trained on a set of labeled training images of a subject in different poses, textures, and background environments. The scheme utilizes metadata stored as 3D models or rendered images of the subject to identify the labeled data and create a classification model. This model can accurately classify a depicted subject in various environments and poses.

Patent/Innovation:

• Machine learning scheme trained on labeled training images of a subject in different poses, textures, and background environments.
• Utilizes metadata stored as 3D models or rendered images of the subject.
• Automatically identifies labeled data to create a classification model.
• Classification model can accurately classify a depicted subject in various environments and poses.

Potential Applications:

• Facial recognition systems for security purposes.
• Virtual reality and augmented reality applications.
• Human-computer interaction and gesture recognition.
• Animation and gaming industries.
• Medical imaging and diagnostics.

Problems Solved:

• Overcomes the challenge of accurately classifying subjects in different poses, textures, and background environments.
• Provides a solution for training machine learning models on labeled data with diverse variations.
• Enables accurate identification and classification of subjects in various real-world scenarios.

Benefits:

• Improved accuracy and reliability in classifying subjects in different environments and poses.
• Enhanced performance of facial recognition systems and other related applications.
• Increased efficiency in human-computer interaction and gesture recognition.
• Enables realistic and immersive experiences in virtual reality and augmented reality.
• Advances medical imaging and diagnostics by accurately identifying subjects in diverse scenarios.

Original Abstract Submitted

A machine learning scheme can be trained on a set of labeled training images of a subject in different poses, with different textures, and with different background environments. The label or marker data of the subject may be stored as metadata to a 3D model of the subject or rendered images of the subject. The machine learning scheme may be implemented as a supervised learning scheme that can automatically identify the labeled data to create a classification model. The classification model can classify a depicted subject in many different environments and arrangements (e.g., poses).