Computer Vision
Tasq.ai's unique computer vision solutions for the creation of high quality datasets for machine learning, enable the process of data labeling at ease.
With Tasq.ai, your computer vision applications will recognize and identify objects at unmatched speed, improving your models’ predictions and confidence levels.
Tasqers Vs Competitors
Type | Progress | Accuracy | Objects | Workers |
---|---|---|---|---|
Tasq.ai | 156 | 198 | ||
Single annotator | 6 | 1 | ||
Internal team | 19 | 6 | ||
BPO | 43 | 30 |
An Iterative Approach to Data Labeling
The Tasq labeling solution is designed from the ground up to scale, without compromising on quality. It’s been the focus of our innovation from day 1, enabling millions of Tasqers to work on your annotation job. We do it with a groundbreaking combination of micro-tasks, instant qualification and quality refinement.
Building an Image Grid
Micro-tasks are all about making complicated requirements into simple ones. We define a workflow for each labeling project. The workflow breaks down the work of image labeling down to simple, easy to understand tasks. This enables hundreds of Tasqers to work on an individual image.
Validating Data to Build Mathematical Confidence in Results
Once an object is annotated and labeled, Tasqers validate the existence of the object in an image or video. Tasqers responsible for validating results are only shown relevant portions of images and asked whether or not the image they are looking at contains the object.
The Tasqers’ multiple judgments are validated, weighted, rated and aggregated into a structured schema of actionable insights.
Creating a Predefined Schema
The workflow also acts as a schema of the resulting annotations, defining a clear structure and relationship between annotations that further improves quality.
Instant Qualification
Tasq.ai has the ability to train a Tasqer in minutes, to perform a micro-task. It relies on sophisticated algorithms to train, qualify, test and monitor digital workers to perform micro-tasks accurately and diligently.
Wisdom of the Crowd
Having an unprecedented amount of workers applied to an image allows us to change the paradigm of labeling from ‘quality sampling’ to ‘quality refinement’. Instead of sampling worker data and projecting quality, we assign each and every micro-task to multiple workers and then collect additional judgments until the required quality metrics are achieved. This creates the first true ‘Wisdom of the Crowd’ solution when every individual label was executed by multiple Tasqers and represents a consensus of opinion.