Furniture Annotation Specialist

Furniture Image Annotation for YOLO Training (20,000+ Images, Pre-Labeled Setup Provided)

Description

We are looking for experienced image annotation specialists to help us prepare training data for a YOLO object detection model.

We already have:
• Label Studio environment fully set up (you will get direct access).
• Pre-defined label taxonomy (all furniture categories are clearly listed).
• Pre-trained YOLO model results (initial bounding boxes are auto-generated).

Your job will be to:
• Review auto-generated bounding boxes for 20,000+ furniture images.
• Correct any inaccurate boxes (move, resize, delete).
• Draw missing boxes for unlabeled objects.
• Ensure that each object is labeled with the correct furniture category.
• Maintain consistent quality and follow our annotation guidelines strictly.

Quality Acceptance Criteria

To ensure high-quality dataset creation, we will perform random QA checks on your work. The project will be accepted only if it meets the following standards:
• Bounding Box Precision
• Boxes must tightly fit the object (no excessive background).
• No truncated boxes unless the object is genuinely cropped in the image.
• No overlapping duplicate boxes for the same object.
• Label Accuracy
• Each object must be labeled with the correct furniture category from our predefined list.
• Mislabel rate must be < 2% in random QA samples.
• Coverage
• All visible furniture objects in the image must be labeled (recall rate > 98%).
• No missing major objects (e.g., sofa, chair, bed).
• Consistency
• Follow the same category usage and box style across all images.
• If you are unsure about a label, flag it for review — do not guess.
• Review Process
• We will perform QA on a random 5% of your annotated data per milestone.
• If error rates exceed the thresholds, the batch must be corrected before payment release.

Requirements
• Proven experience with image annotation (YOLO, COCO, Pascal VOC, etc.).
• Familiarity with Label Studio (or similar tools).
• Understanding of bounding box quality standards (tight fit, no overlaps, correct class).
• Attention to detail – quality is more important than speed.
• Ability to work with large datasets and meet deadlines.

Deliverables
• Fully annotated dataset exported from Label Studio in YOLO format.
• Clean, consistent labels ready for training the next YOLO iteration.

Project Size & Payment
• 20,000+ images, mostly furniture items.
• Payment can be milestone-based (e.g. per 5,000 images) or per hour — negotiable based on quality and speed.

Preferred Candidates
• Individuals or teams with prior object detection dataset creation experience.
• Can start immediately and dedicate time to finish the project in a reasonable timeframe.