AI 3D Face Model for Low-Light Video Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing laptop computers and devices face challenges with low video quality due to suboptimal image sensor characteristics, particularly in low lighting conditions, leading to adverse effects like sub-optimal frame rate and image blurriness.
Innovation Solution
A device with a processor and storage that facilitates video conferencing by identifying lighting conditions and using AI/3D models to generate or modify video streams, ensuring improved image quality by transmitting either real-time video or enhanced 3D representations based on lighting criteria.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If exposure time is extended to improve image quality in low lighting conditions, then image quality is improved, but frame rate becomes sub-optimal and image blurriness increases
Solution Approach 1:
The system performs preliminary actions by capturing training images during well-lit conditions to build a 3D model and AI model of the user's face before the actual video conference. This pre-processing enables the system to generate high-quality video representations in real-time during low-light conditions without compromising frame rate, as the computationally intensive model training is completed in advance.
Solution Approach 2:
The system creates a 3D digital copy and AI model representation of the user's face based on training images captured in well-lit conditions. This digital copy can be rendered and manipulated to generate realistic video representations during low-light conditions, replacing the need to use actual low-quality captured images and eliminating the trade-off between exposure time and frame rate.
2Illumination intensity
If exposure time is extended to improve image quality in low lighting conditions, then image quality is improved, but image blurriness increases
Solution Approach 1:
The system performs preliminary actions by capturing training images during well-lit conditions to build a 3D model and AI model of the user's face before the actual video conference. This pre-processing enables the system to generate high-quality video representations in real-time during low-light conditions without compromising frame rate, as the computationally intensive model training is completed in advance.
Solution Approach 2:
The system creates a 3D digital copy and AI model representation of the user's face based on training images captured in well-lit conditions. This digital copy can be rendered and manipulated to generate realistic video representations during low-light conditions, replacing the need to use actual low-quality captured images and eliminating the trade-off between exposure time and frame rate.
Data Source
AI summary
In one aspect, a first device includes at least one processor and storage with instructions executable by the processor to receive input from a camera and to identify, based on the input from the camera, a first lighting condition of a scene showing a user. Responsive to the first lighting condition satisfying a first criterion, the instructions are executable to transmit a first video stream showing the user's face as captured by the camera to a second device. Responsive to the first lighting condition not satisfying the first criterion, the instructions are executable to use a model to generate a 3D representation of the user's face according to one or more parameters related to the user and to transmit a second video stream to the second device that shows the 3D representation. The model is also trained under lighting conditions satisfying one or more criteria.