NeuralMarker platform allows you to bring your own data and team of labelers to the platform and allow the team to label the data for AI use-cases very efficiently. In this article, we shall learn the key concepts of the tool. Here are the key building blocks:
1. Dataset: Dataset is a collection of images or videos that we have identified to be labeled and used for training an AI model.
2. Category: Categories are just names of objects that we shall assign while labeling.
3. Category_type: Category types refer to the kind of annotations that labelers would mark. There can be many category types:
a. 
Rectangle: Labelers can mark a rectangle around the objects of interest and assign them names.
b. 
Polygon: Labelers can mark a polygon around the objects of interest and assign them names.
c. 
OCR: Labelers can mark a rectangle or polygon around the objects of interest and specify the text written within the polygon or rectangle.
d. 
KeyPoints: Labelers can mark points or curves on the images to specify the object of interest.
e. 
Segmentation: Labelers can mark each pixel of the objects of interest and assign them names.
f. 
Classification: Labelers can specify names for the images.
4. Pre-trained Models:
In order to speedup annotations, NeuralMarker has many in-built AI models that can be used to mark whole datasets before labelers start their work. The data scientists in your organization can also add their own models to help annotation work. These custom models would only be visible to your organization along with public pre-trained AI models.