Self-Confident: Online Learning for Detecting Depth Sensors Failures
Introduction
The invention relates to a method and a sensor system, designed in particular for determining the confidence of disparity maps inferred by a stereo algorithm or a network through a neural network capable of self-adapting.

Technical features
There are in the market several systems for acquiring images in 3D, in order to determine the depth of an image. Stereo is one of the most popular strategies used to accurately perceive the 3D structure of the scene, through synchronized cameras and several algorithms.
The invention relates to a method and a sensor system, designed in particular for determining the confidence of disparity maps inferred by a stereo algorithm or a network through a neural network capable of self-adapting, but which can be used for any type of image acquisition system, in which it is necessary to estimate the confidence, thus determining the level of certainty or uncertainty of each pixel of said image. Estimation of self-adaptive confidence could be highly desirable for many practical application, as smartphones that are equipped with multiple cameras and stereo algorithms deployed for augmented reality or other applications in unpredictable environments.
Possible Applications
- Autonomous driving systems;
- Computer Vision;
- Robotics;
- Augmented Reality;
- 3D reconstruction.
Advantages
- Method for self-adapting a confidence measure unconstrained to the stereo system deployed;
- High reliability;
- Easy to implement;
- Competitive in terms of costs.