The human factor in the ML training loop is essential to obtain high quality labels and model results.
Ango Hub’s AI Assistance features, instead of replacing the human element, are designed to complement it by removing repeating tasks and making it easier for the annotator to express their natural intuition.
At Ango AI, our sole goal is delivering great quality labels at scale. And to achieve this goal, labeling efficiency and quality are the two most important factors.
AI Assistance tools help in achieving exactly this, increasing labeling efficiency and quality, all while maintaining the human element that distinguishes our top-quality annotations.
Many platforms in the data labeling category offer highly specialized AI assistance tools, tailored to speeding up labeling in certain well-known scenarios.
We want to develop AI Assistance tools that can help everyone, not just a few. This is why our tools are field-agnostic: no matter your project, you will find them useful.
Here’s some of them.
FrameCut is our agnostic polygon/segmentation tool. It makes the process of annotating easier by several degrees of magnitude, letting deep learning do most of the heavy lifting to obtain a pixel-perfect mask for an object.
It’s simple to use: just draw a box around the object you’d like to label and place one or two clicks to refine the final polygon.
FrameCut supports segmentation and detection, it works for an extremely wide range of objects, and delivers incredibly high efficiency.
AutoDetect is one of those tools that feel like magic the first time you try them. It allows for detection of common objects within a complete image.
The network this tool relies on has been trained on a massive dataset to ensure it can comfortably detect a plethora of different objects.
To use it, simply map your dataset’s classes to the classes of AutoDetect and click on a single button. That’s all.
For common objects, AutoDetect is the fastest AI tool in the arsenal.
Optical Character Recognition (OCR) is a must-have, and of course Ango Hub includes it as well.
With a single click, it detects, transcribes, and localizes the text on an image through bounding boxes. All in the order of mere seconds, for images that would manually take minutes to annotate.
Our OCR implementation works with a myriad of different languages and drastically simplifies document labeling.
Smart Scissors is an interactable, agnostic tool for encapsulating objects within a polygon. It uses cutting-edge computer vision methods to detect object boundaries based on mouse movement.
Using it is intuitive. Start with a point on the boundary of the object, and start dragging the cursor around the object’s shape. Smart Scissors will make sure you draw a tight polygon around your object, no matter you actual accuracy.