Polygon Segmentation of Single-Object Images

About OpenTrain

OpenTrain is the #1 platform for building careers in AI training and data labeling. We connect contributors with projects that help teach and refine modern AI systems while offering flexible, remote work opportunities.

About AI Training Work

AI training (also called data labeling or annotation) is the human side of building artificial intelligence: people create and correct examples that models learn from. Contributors often work remotely, choose flexible hours, and directly influence how state-of-the-art models behave.

The Role

You will perform accurate polygon segmentation of a single target class in images using our hosted CVAT instance. The project uses a growing dataset of roughly 18,000 images today and will continue to expand; labeling is continuous as new data arrives.

• Label type: segmentation (polygon) for one target object per image.
• Tool: CVAT (hosted instance provided by the project).
• Method: polygon annotation; you may use Segment Anything (SAM) to propose masks but must manually correct them.
• Dataset size: ~18,000 images and increasing; workflow is continuous.

What You'll Do

Complete polygon annotations for each image following the project's labeling guidelines. Use SAM suggestions to speed up work but inspect and correct every mask manually to ensure accuracy.

• Open tasks in the hosted CVAT instance and follow per-task instructions.
• Draw or edit polygons tightly around the target object; ensure masks do not include background or other classes.
• Validate and refine SAM-generated masks where used; never submit without manual review.
• Submit completed labels and move on to the next task as data becomes available.

Requirements

This is an entry-level contractor role with no additional formal requirements listed. You should be comfortable using a computer and web tools and able to work reliably with attention to detail.

• No prior annotation experience required — suitable for beginners.
• Access to a reliable internet connection and a computer capable of running a browser with CVAT.
• Willingness to follow detailed labeling instructions and quality standards.

Who Should Apply

Apply if you want flexible, remote work contributing to computer-vision model training. Candidates who are patient, detail-oriented, and comfortable with polygon editing tools will succeed in this role.

• Ideal for people seeking part-time or flexible contractor work.
• Good fit for those curious about AI and hands-on model training tasks.
• Open to applicants worldwide.

Compensation & Schedule

This project pays per labeled image at a rate of $0.04 USD per label. Work is contracted and task-based; there is no hourly salary listed and you control how many images you process.

• Payment type: pay-per-label at $0.04 USD per image.
• Employment type: contractor; schedule is flexible and remote.
• Workload: continuous labeling as new data is added; you choose how much to take on.

How It Works

Create an OpenTrain account, build your profile, and apply through the platform. Once accepted you will get access to the hosted CVAT instance and project-specific instructions; then you can begin labeling and submit work as tasks are completed.

• Sign up on OpenTrain and complete any required onboarding steps shown in your profile.
• Access the project through the platform, open CVAT tasks, and follow the annotation guidelines provided.
• Continue labeling as new batches are released; monitor the project for updates and additional instructions.