Button Annotation for Product Photos (Bounding Box & Polygon)

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We connect people to cutting-edge projects that help teach AI systems how to see, read, and understand the world.

Working through OpenTrain lets you discover projects, build a profile, and apply in minutes — all with flexible, remote-friendly work that suits part-time contributors and people starting out in the industry.

About AI training and image annotation

AI training (also called data labeling or annotation) is the human work behind machine perception. Annotators create the examples modern computer-vision models use to learn — drawing boxes, polygons, and labels on images so models can recognize objects precisely.

This project focuses on button/icon detection in photos: your annotations will directly help machine-learning systems learn to spot interface elements and visual buttons under real-world conditions like tilt, perspective, and occlusion.

The role

You will label buttons and related UI elements inside photos. Each image contains 5 or 6 buttons or objects (an icon with text beneath). Because photos are taken at angles, polygon annotations are often needed to capture the true shape of each element.

This is an entry-level contractor role that accepts contributors worldwide. Work remotely and set your own hours — label at your own pace.

• Data type: images containing 5–6 buttons/objects each
• Label types required: bounding boxes and polygons
• Labeling tool: Label Studio
• Employment: contractor, worldwide contributors accepted
• Rate: $0.05 per labeled object (per label)

What you'll do day-to-day

Follow a provided guideline to find each button or object in a photo and annotate it precisely with the requested label type (bounding box or polygon). Use polygons when the element is tilted or has an irregular shape due to perspective.

Submit accurate, high-quality annotations on Label Studio and correct any issues flagged by reviewers. Maintain consistent labeling across images to improve model training quality.

• Identify the icon + text grouping for each button in the image
• Draw bounding boxes for roughly rectangular elements and polygons for angled or irregular shapes
• Apply consistent labels according to the project guidelines
• Review and correct annotations when requested

Requirements

This is an entry-level project: no prior professional annotation experience is required. You must be comfortable working with images and following detailed labeling instructions.

You will use Label Studio to complete tasks; familiarity with the tool is helpful but not mandatory. Attention to detail and the ability to maintain consistent quality across many images are essential.

• Entry-level — prior annotation experience not required
• Willingness to use Label Studio and follow project guidelines
• Ability to annotate polygons and bounding boxes accurately
• Access to a computer or device capable of running a web-based labeling tool

How pay and work are structured

This project pays per labeled object at a rate of $0.05 per label. You will be paid based on the number of valid annotations you submit and that pass quality checks.

Work is contract-based and flexible — you choose when and how much you annotate. Exact payment frequency and payout method will be shared when you join the project through the OpenTrain platform.

• Compensation: $0.05 per labeled object (per label)
• Contractor work with flexible hours
• Worldwide contributors are welcome

How to apply

Create an OpenTrain account (free) and apply to this project from the listing. You will receive the project guidelines and access to Label Studio once accepted.

If you enjoy careful, visual work and want to help train real computer-vision systems, apply now — contributors often scale up quickly as they build experience and reputation.