Looking for a Side Hustle? Become a Computer Vision Annotator

In the midst of trending city/state/country lockdowns and social distancing measures due to COVID-19, we at Voxel51 want to highlight a vital role to the Computer Vision industry: the data annotator.

Although Artificial Intelligence (AI) has yet to achieve human-level understanding, the Computer Vision (CV) space is seeing rapid advancement towards providing machines with sight using Machine Learning (ML) and, in particular, Deep Neural Networks (DNNs). CV models, which extract insights from camera streams, images, and videos, continue to improve as more quality, curated data is pushed through training pipelines. Current industry-leading CV training methods still depend on human intervention to deliver high-quality annotations for training, which results in a huge demand for… human annotators! In a Lex Fridman interview of Cristos Goodrow, VP of Engineering at Google, Cristos mentions the importance of humans-in-the-loop for training ML models, and we couldn’t agree more.

Annotation jobs, many of which are remote with flexible hours, might just be the perfect job in this new age of quarantine and social distancing. Is data annotation the right gig for you? Let’s address some questions that may be crossing your mind.

Do I have what it takes?

Although just about anyone can become an annotator, the job itself is not for everyone. In our experience, successful annotators demonstrate the following skills:

• Good spatial reasoning skills
• Ability to work independently in focused blocks of time to meet deadlines
• Enjoy performing detail-oriented and quality-focused work at speed
• Proficient with basic computer skills and navigation (i.e. using a mouse/touchpad, shortcuts, hotkeys, etc.)
• Experience with a Computer Vision annotation tool (see below for a screenshot from our internal annotation tool)

What does compensation look like?

As with all potential jobs, you’ll want to understand the compensation structure. For annotators, it will depend on the employment relationship. A full time employee may be compensated differently than a freelancer or contractor. If you’re looking to do annotation as a side hustle, you may get paid per object, by the hour, or by the project. In general, compensation closely follows two factors: quality and efficiency. The bottom line is, the more annotations you can produce at a high quality, the greater your earning potential.

What is the day-in-the-life of an annotator?

In general, you can expect the following workflow when it comes to performing annotations:

1. Get briefed on the goals and what needs to be annotated (ontology, protocols, practices, schema, etc.)
2. Fire up a web browser, login to the annotation tool, and accept an assigned project or task
3. Annotate data. Save work. Take breaks. Repeat.
4. Submit your annotation tasks
5. Look for feedback from the quality assurance team; refine tasks that were flagged for improvement; and re-submit
6. Get paid!

What are the perks?

Annotation may not be the right fit for everyone, but here are the top reasons we’ve heard for why people want to become annotators:

• Work from home for companies around the world
• Set your own schedule
• Gain exposure to interesting data
• And our favorite — help advance technologies at the frontier of human progress, which will define our evolution out of the information age

How can I learn more?

There are too many resources to fully list, but here’s some guidance on getting started:

1. Learn about the various types of annotations
2. Get a feel for annotation tools
3. Familiarize yourself with concepts related to high-quality annotations such as: classification, occlusion, object tracking, truncation, data diversity, data bias, and ontology

Where are the jobs?

Although you can probably get a good feel for what large organizations look for in annotators on LinkedIn, ZipRecruiter, or Glassdoor, most part-time jobs can be found in sites or services like these:

• Freelance on crowd-sourcing sites like Fiverr and check out how others present themselves
• Take on tasks from Requesters on Amazon Mechanical Turk
• Do a Google search like this for companies that provide CV/AI capabilities
• Lastly, at Voxel51, we have a community of people seeking great annotators. Come join us on Slack to chat with them!

At this stage of the CV/AI revolution, high-quality annotated data, images, and videos are a critical need for companies who want to be first movers in the space. Joining this effort as an annotator can be a very rewarding experience, and you can easily get started today from the comfort of your home, despite the COVID-19 pandemic. It’s not difficult to become an annotator and the effort is relatively light. Claim your piece of the AI development legacy today!

About Voxel51

High-quality, intentionally-curated data is everything when it comes to training great CV models. At Voxel51, we have over 25 years of CV experience and think deeply about achieving state-of-the-art outcomes for the CV community. To this end, we developed FiftyOne, a tool designed for machine learning engineers and scientists to rapidly experiment with and debug their datasets and models. Time is of the essence in bringing your AI solution to life. If you want to accelerate and strengthen your models and data pipelines to make it happen, let’s chat! Join us on Slack and check out FiftyOne!