Virtual reality is a box of old children's toys that live as long as a child's imagination lasts. Then the box becomes too small to contain the entire world, and it is repurposed for tools. Augmented reality is a tool for changing the world, starting with exploring it.
How is data labeling via AR better than the current way of labeling data? Let's consider two cases: for road obstacle detection, and for medical tumor recognition. In both cases, data labelers must label a large number of objects in photographs. This database, then converted into a specific format, serves as a pattern for recognition. Such a system is weakly protected against errors.
In the case of AR, the data labeler directly interacts with an object on the road or a sample of malignant tissue. Here, both logic and subjective factors (impressions) are involved, providing the neural network with additional information for processing. Such a system should be more error-proof, as first impressions are more trustworthy.
Now the person labeling the data spends less time on unproductive work and becomes more of an expert in each situation.