collective mind AG
Contact at the AI Innovation Center
Chandan Kumar
Intuitive label editor for AI applications
Initial situation
Collective mind is developing a platform specially tailored to industrial use cases for the use of artificial intelligence (AI) and machine learning, which accelerates and improves the implementation of customer projects. A key component is the label editor, which enables the curation and labeling of training data for AI models. Currently, labeling data, especially image data, is a time-consuming and resource-intensive process that is largely carried out manually. This leads to high costs and can delay the introduction of AI-based solutions.
In order to better understand the needs and challenges of users, interviews were conducted with data labelers at the start of the project. These discussions revealed both positive and negative aspects of current external solutions. It became clear that although the existing labeling tools fulfill their purpose and support the manual process to some extent, small details in the workflow often interfere, are cumbersome or are missing altogether.
Solution idea
Our solution idea addresses the identified challenges with an intuitive labeling editor. This guides users through the labeling process with the help of structured workflows and aims to significantly reduce effort and increase the user-friendliness of the editor. Support and automation functions relieve users of time-consuming, repetitive tasks. For example, the basic model for image segmentation ("Segment Anything") or the "Visual Prompting" approach enable model-based labeling, which makes the process more efficient.
Benefit
The implementation of an efficient data labeling system in the AI platform offers considerable added value. Companies are able to work with larger training data sets than before, which can increase the performance of the AI models. The integrated support functions also reduce errors in the labelling process and minimize monotonous work steps. These optimizations reduce costs and increase efficiency, which also benefits existing processes.
The fact that data labeling will be less complex in future will also help to speed up the development of AI models.
This can strengthen the competitiveness of customers from the value-adding industry by enabling them to realize innovative solutions more quickly and establish them on the market.
Implementation of the AI application
The intuitive labeling editor is based on an expandable concept that can be flexibly adapted to different requirements and extensions. A structural organization into interaction states ensures a clear and concise user interface. The comprehensible and context-sensitive arrangement of the functional areas promotes intuitive operation and efficient work processes.