Vision AI Brains
There are three main types of Brains that are used in Vision AI systems: Recognition, Detection, and Segmentation.
Recognition Brains analyze a whole image or video frame and determine if the content matches a class on which the Brain has been trained. These types of Brains are best used for recognizing objects (e.g. "dog") or scenes (e.g."beach").
Similar to Recognition, Anomaly Recognition Brains examine the whole image. An Anomaly Recognition Brain is trained to recognize what "good" data looks like and then identify any defect or variation to find images that contain anomalies. These Brains are best used in use cases such as manufacturing, where images are very consistent from one to the next and there may be ample quantities of "good" images but few that contain instances of defects or anomalies.
Detection Brains can find multiple objects in an image—of the same class or of multiple classes—and will show the location of the detected objects with rectangular bounding boxes.
Segmentation Brains can find and locate multiple objects within an image or video, and show their location with pixel-level precision.
Brain Builder Capabilities
Brain Builder can currently be used to train and deploy Recognition Brains using Neurala's patented online training technology, and to annotate images and videos for Segmentation Brains and Detection Brains to be used in offline training processes.
Organizing your Data
Brain Builder uses Projects and Datasets to organize your data into logical groupings for easy management and straightforward workflows.
The structure between Projects, Datasets, and Data is simple:
Your Brain Builder environment can have multiple Projects.
Each Project can contain multiple Datasets.
Each Dataset can be used for one type of Brain—either Recognrition Brain training, Detection annotation, or Segmentation annotation—and can contain many images and videos.
Making the Most of Projects and Datasets
When organizing your Brain Builder environment, you can:
Use Projects to keep different teams' and stakeholders' work separate (for Enterprise customers).
Use Datasets to create multiple Recognition Brains using different collections of images.
Use Datasets to help your taggers focus on a single type of object when they tag by grouping content with the same classes.
Use Datasets to organize your data in date-based batches for easy tracking and exporting when uploading additional data over time.
Assign users to only view specific Datasets so you can preserve data privacy and limit access to only those who need it. (For Enterprise customers. See Administration for details.)
It's that Simple!
Now you should understand the basics of the types of Brains you can build in Brain Builder and how you can use Projects and Datasets to effectively organize your work.
Next: Creating a Project >>