Polygon Segmentation Specialist

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We help people start and grow careers teaching AI by connecting contributors with projects, tools, and training — creating flexible, remote work opportunities in a fast-growing field.

OpenTrain AI (the hiring organization for this role) runs this project directly. Creating an OpenTrain account is free and lets you apply, track tasks, and build a profile that can lead to more work across the industry.

About AI training and why it matters

AI training (also called data labeling or annotation) is the human work that teaches machine perception: people annotate images, audio, text, and more so models learn correct behavior. This work is often remote, flexible, and accessible to contributors without formal credentials.

Working on image segmentation means you directly shape how computer-vision systems see and separate objects — a hands-on way to be part of cutting-edge AI development.

The Role

You will generate polygon segmentation masks for objects in images using CVAT alongside the Segment Anything tool. Objects larger than a specified size threshold must be outlined with polygons that separate individual items when multiple objects appear in a frame.

This is an entry-level, contractor, part-time role with flexible hours (under 20 hours per week). Work is remote and open worldwide.

• Label type: Polygon segmentation on image data
• Labeling software: CVAT, used together with the Segment Anything interactor
• Employment: Contractor, part-time, remote

What you'll do day-to-day

Perform careful image-by-image polygon annotations, ensuring each qualifying object is individually outlined and separated from nearby objects. Use the Segment Anything interactor to speed up selection, then refine polygons in CVAT.

Follow project size thresholds and annotation guidelines (we will provide instructions). Maintain consistent quality and meet per-session speed and accuracy targets.

• Create precise polygon outlines for objects above a size threshold
• Distinguish and separate individual objects when multiple items appear
• Use Segment Anything to propose masks and edit polygons in CVAT
• Follow project-specific annotation rules and quality standards

Requirements

No formal credentials are required — this role is open to entry-level contributors who can learn the tools and follow guidelines. Familiarity with CVAT and the Segment Anything interactor is advantageous but not mandatory.

Attention to detail and the ability to reliably distinguish unique objects in cluttered frames are essential. You must have a computer, stable internet, and the willingness to complete a short trial assignment.

• Willingness to learn CVAT and the Segment Anything tool
• Strong visual attention to detail and object separation skills
• Access to a computer and stable internet connection
• Able to work under 20 hours per week and accept a paid trial

Compensation, hours, and logistics

This contract role pays per hour at a rate between USD $4–$6/hour (hourlyRate set at $5/hr). Work is part-time and flexible: expect less than 20 hours per week unless otherwise arranged.

We will begin with a small paid trial to assess speed and accuracy. Contributors who meet our requirements may receive additional assignments.

• Pay: $4–$6 per hour (hourly pay per contractor agreement)
• Time: Less than 20 hours per week; schedule you control
• Trial: Small paid trial at project start; successful results can lead to more work

How the hiring process works

Apply through your OpenTrain profile and accept the short trial task. Complete the trial using the provided instructions and tools. We will review your annotations for accuracy and speed.

If you meet the quality and speed benchmarks, we'll offer additional contract work. Feedback and simple onboarding materials will be provided so you can ramp up quickly.

• Step 1: Apply and accept the paid trial assignment
• Step 2: Complete annotations in CVAT using Segment Anything where helpful
• Step 3: Receive feedback; qualified contributors may get more work