Collection and processing
Know-how for Curating 200M+ Datasets
Introducing Datumo's systematic data building infrastructure and technology, from recruitment to task execution and verification.
Our Labelers
245K users are registered in Cash Mission, Datumo’s crowd-sourcing platform(last updated: January, 2023).
Using both web and app version of the platform, crowd-workers can participate in projects whenever and wherever, as much as they would like.
More and more projects are requiring higher complexity, security, and confidentiality.
It is also important to have qualified workers for particular projects.
Datumo is equipped with both human resources and facilities, including in-house voice recording booths and annotators’ workspaces.
Technology
• Ensure the outline of target object lies between bounding box and guide box, which enables accurate and consistent bounding box annotation
• Ensure the margin between bounding box and object outline is consistent by forming a “guide box”
• Guide box is formed based on the IoU(Intersection over Union) set according to bounding box
* Patent number 10-2176458
• Save approx. 50% of time consumed in labeling, compared to manual labeling
• Semi-automation enabled by small amount(400) of data, regardless of domain, using our own data augmentation technology
Work
User Interface
Cash Mission develops and provides interface that improves the work efficiency of crowd-workers. Appropriate UI lowers the difficulty of collecting data such as voice, text, images, and more, and also saves time.
The project managers constantly communicate with annotators, make their own training materials, and manage the annotators.
Datumo goes beyond generic training of data annotators.
Inspection
