Get Started with Refonte.Ai
Pro
Nucleus
Key Concepts & Definitions
The first step in using Refonte.Ai to obtain high-quality ground truth data is to start a project. You will upload data into a project and create tasks, which are segments of data that need to be labeled. To start the task grouping for labeling, separate batches may be formed. Each task will adhere to the same project-level taxonomy.
You can create tasks using our UI or by submitting an API call after your data is hosted in a manner that allows Refonte.Ai to access it. Your tasks' statuses will be "pending" once you have launched a batch of tasks for labeling.
- Customers of Refonte.Ai Rapid can often anticipate receiving production batches in a few days or up to a week, and calibration batches in a few hours or up to a day. Be aware that more intricate labeling use cases could require more time. Please contact us on Intercom if you have any questions concerning turnaround time or would want a faster throughput.
- Customers of Refonte.Ai Studio have the ability to manage the speed at which they receive jobs back because they are using their own annotation team. In order to control turnaround time and quality, studio clients collaborate directly with their own annotators.
- Refonte.Ai Pro clients can anticipate receiving these assignments back in accordance with the delivery timeline that we have agreed upon with you. We can accommodate dynamic, very large volumes that are tailored to your requirements.
Once a task has been labeled, you'll see the task status move to be “completed.” The task will now have a JSON response associated with it that you can download via our platform.
The online application allows you to export a large number of jobs in bulk over a filterable range or download the response for a specific task. With the use of a task ID, you can programmatically retrieve tasks from our list of APIs, or we can list all tasks that match your customisable filter criteria. Last but not least, we fully support callbacks, enabling complete programmatic access to your labeled data as jobs are completed, tagged with an error, or undergo other actions.
Task
A task is a discrete piece of work that needs to be completed. The task and the data that needs to be tagged are mapped one to one. For instance, each image, video, or text that needs to be tagged will have its own job, each with an own Refonte-generated ID. Please go to our API reference to learn how to create a task using our API.
Project
Similar jobs within a project can be grouped together according to the use case and instructions. The annotation guidelines and instructions will be the same for every task.
One particular annotation use case, linked to a task type in our API documentation, is the focus of a project. For each use case, you may have more than one project.
For instance, you may have two projects: one for image annotation and the other for scene classification.
To keep things organized, each task is connected to a certain project. Please refer to our API reference to establish a project utilizing our API.
Batch
On Refonte.Ai Rapid: You can send batches of data for labeling to the Refonte.Ai workforce within your projects. You can read more about the three different kinds of batches on Rapid—self-label, calibration, and production batches—here.
On Refonte.Ai Studio: You can launch batches of data to be annotated by your own team within your projects. Although all of the batches are regular production batches, you are free to use them however you see fit (for example, label the batches yourself, use them for large refonte production pipelines, or use them as an experimental batch with your annotators).
On Refonte.Ai Pro: Batches are an alternative way to further segregate work inside a project for big volume projects. Batches, for instance, can be used to identify which jobs were a part of a weekly submission or to link tasks to certain datasets you use internally.
Taxonomy
Defined at the project level, a taxonomy is a set of labels and the data that goes along with them. Every label is referred to as an annotation. Box, polygon, point, ellipse, cuboid, event, text response, list selection, tree selection, date, linear scale, and ranking are among the available annotations. There might be classes of annotations (i.e., various kinds of annotations), global attributes (i.e., details about the entire task), and annotation attributes (i.e., details related to a particular annotation) inside a taxonomy. Additionally, link attributes—that is, the connections between two annotations—can be created.
Example: As an illustration, one use case may entail indicating the total number of cats and dogs in an image by drawing boxes around each one. We want to be able to tell whether each cat is asleep or not. If applicable, we would like to note which cat each dog is staring at.
Two classes of box annotations—one for cats and one for dogs—would be included in our taxonomy. To correlate each box drawn around a cat with whether or not it is sleeping, we would define an annotation attribute of "sleeping or not sleeping" within the cat class. To attach a dog box to a cat box and denote a "looking at" relationship, we would define a link property within the dog class. Lastly, we would design a global property that would ask the labeler how many dogs and cats there are in the image overall.
Updated about 1 month ago