Semantic Segmentation of a Single Object per Image with Segment Anything

About OpenTrain

OpenTrain connects people with data-labeling and AI-training work. We aggregate annotation projects so contributors can discover opportunities in one place and apply quickly.

Creating an OpenTrain account is free and applications take only a few minutes. This listing is for a contract project managed by OpenTrain AI using our hosted tooling.

• Centralized job board for data-labeling and AI-training roles.
• Free to join and quick to apply.

About AI training and this kind of work

AI training (also called data labeling or annotation) is the human contribution that teaches models to see, speak, and understand. Tasks range from drawing bounding boxes and polygons to transcribing audio and rating model outputs.

These projects are commonly remote and entry-level friendly; many let contributors choose when they work. Your annotations directly help computer vision models learn accurate object boundaries.

• Human-labeled examples make AI systems accurate and useful.
• Work is often remote, flexible, and accessible to entry-level contributors.

The role

You will create precise polygon segmentations for a single target object class in images using our hosted CVAT instance.

This is a contract, pay-per-label role open worldwide and suitable for entry-level annotators. The dataset currently contains about 18,000 images and will grow over time.

• Data type: images (single object per image).
• Label type: segmentation (polygon annotations).
• Project scale: ~18,000 images and expanding.
• Employment type: contractor, worldwide.

What you'll do

Label images by drawing accurate polygon masks around the single specified object in each image. You may use Segment Anything Model (SAM) to generate initial masks, but you must manually correct and refine every mask to meet quality standards.

Work is continuous: new images are added over time and you can pick up tasks as they become available through the hosted CVAT platform.

• Use CVAT (hosted instance) to open, annotate, and submit images.
• Apply polygon annotation tools to produce precise object outlines.
• Use SAM suggestions when allowed, but always verify and correct outputs.
• Follow visual guidelines for consistent segmentation across the dataset.

Compensation and workflow

This project pays on a per-label basis: $0.04 USD for each completed image annotation. You will be paid as a contractor per the platform's payment process.

Because labeling is continuous, you can work when tasks are available. There is no fixed hourly commitment; throughput determines earnings.

• Pay rate: $0.04 USD per labeled image.
• Payment type: pay-per-label (contractor).
• Schedule: continuous labeling as new data is added; work on tasks as available.

Requirements

This is an entry-level role and requires no prior experience beyond the ability to follow instructions and pay attention to visual detail.

You must be able to use a web browser to access CVAT and perform polygon annotations accurately.

• Experience level: entry level (no previous annotation experience required).
• Tools: ability to use web-based CVAT (hosted instance provided).
• Technique: draw polygon masks; SAM may be used but manual correction is required.

Who should apply and how it works

Apply if you want remote, flexible annotation work and are comfortable drawing polygons around objects in images. This role suits people new to AI labeling who want steady, on-demand tasks.

To apply, create an OpenTrain account if you don’t have one, submit via the listing, and follow onboarding instructions to access the hosted CVAT instance and project guidelines.

• Suitable for: beginners, students, part-time contributors, and anyone comfortable with visual labeling.
• Onboarding: accept the contract, get access to CVAT, review the annotation guide, and start labeling.