We use cookies to enhance your browsing experience and analyze our traffic. By clicking "Accept All", you consent to our use of cookies.. View our Privacy Policy for more information.
Your browser (Internet Explorer) is out of date. Please download one of these up-to-date, free and excellent browsers:
For more security speed and comfort.
The download is safe from the vendor's official website.


CVAT and seamless integration

Charlie Child
Jan 7, 2021

CVAT and seamless integration

2 min read

HUMAN Protocol establishes decentralized, global marketplaces for human labor, in which everything from the request for work, the evaluation of answer quality, and the delivery of results can be executed by software.

Machine Learning practitioners have a limitless demand for human annotation of raw data. Raw data must be broken down and labelled (or ‘annotated’) by a human for it to be ready for, and improve the quality of, Machine Learning. For example, a drone taken image of a lake is raw data; a human then labels a boat, a canoe, and a dock within this image. The labelled data is primed for machine learning products, because these ‘labels’ can be used to teach machines what each of these objects looks like.

Across the world, a wide variety of data-annotation tools are used by millions of people. HUMAN Protocol lets both requesters of work and workers bring the tools they prefer. This post reviews the process of bringing one such tool, CVAT, onto the Protocol. We hope this serves as a blueprint, showing just how easy it is to expand the ecosystem with seamless integration of other tools.

What is CVAT?

CVAT is a free, open-source annotation tool, created by Intel. CVAT presents human workers with an image, within which they can use a variety of labelling tools to mark up different objects. This can be anything from dragging a rectangle or ‘bounding box’ across a truck, to a freehand shape around a less regular object, such as a human walking across a street. Furthermore, CVAT supports a variety of automated labelling tools. Using a variety of deep learning models, CVAT assists the worker in the segmentation process, facilitating faster labelling work.

What we have done: a seamless way to integrate

The HMT-Escrow tool is a Javascript, Solidity, and Python 3 library used to launch labelling jobs onto HUMAN Protocol.

The job description is sent to a HUMAN “Exchange” server, which handles interaction between the Protocol and our open source extension of CVAT. HUMAN Protocol is engineered to create these unique Exchanges, which manage the distribution of tasks. In this case, the Exchange parses and converts the new jobs into a data blob which CVAT can understand. The data is then distributed to available and appropriate CVAT users connected to the Exchange.

Once the workers submit their completed work, the data is sent back to the Exchange. The Exchange sends results on to a Recording Oracle, which gives an initial evaluation and periodically aggregates results for final evaluation by a Reputation Oracle, which ultimately initiates a payout from the HMT-Escrow contracts.

Implications: CVAT and beyond

The work we have done extends beyond CVAT.

Primarily, companies with specific image and video labelling needs will be able to use this platform to directly access skilled worker pools that will provide data, labelled to a specified precision; it also means that websites and data-labelling companies running CVAT will be able to loop into HUMAN Protocol.

This is a single use case. However, the technological framework we have built demonstrates how companies can plug in their own labelling software, allowing for a seamless and wide-spread adoption of HUMAN Protocol.

For the latest updates on HUMAN Protocol, follow us on Twitter or join our community Telegram channel.

Legal Disclaimer

The HUMAN Protocol Foundation makes no representation, warranty, or undertaking, express or implied, as to the accuracy, reliability, completeness, or reasonableness of the information contained here. Any assumptions, opinions, and estimations expressed constitute the HUMAN Protocol Foundation’s judgment as of the time of publishing and are subject to change without notice. Any projection contained within the information presented here is based on a number of assumptions, and there can be no guarantee that any projected outcomes will be achieved.

Guest post