VS Robotics: Elementary Multimodal Data Markup Platform

The main articles are:

• Big Data
• Data labeling

2024: Start of auto-marking system

The Elementary platform team (project) VS Robotics has launched an auto-marking system neural network with models, which makes it possible to significantly facilitate subsequent markup data AI by trainers. Significant acceleration was achieved in the most routine tasks, resource-intensive for humans, but easily and quickly solved by fundamental neuromodels. VS Robotics reported this on October 21, 2024.

Elementary is a universal platform for multimodal markup of a large array of any complexity of data, which will later be used to improve artificial intelligence and machine learning algorithms. A large number of regular AI trainers mark up huge amounts of data every day, for example, the ethical aspects of the work of the created voice models, compare generative images or translate audio into text.

The auto-tagging system, which includes a set of neural networks and LLM (large language models) for solving specific problems, made it possible to speed up the process of marking up data by 3 times. This became possible thanks to the preliminary processing of the datacet and its subsequent adjustment by people, which reduced the need for completely manual work.

"The use of neural network technologies not only accelerated the markup process on the platform, but also allowed our team to focus on more important tasks, such as improving the quality and accuracy of data. This allows us to use resources more efficiently, and customers to receive more accurate and useful data for the development of their projects in the field of artificial intelligence, "comments Maxim Koloskov, head of VS Robotics.