20k NSFW Image Annotations

I have a batch of roughly 20,000 NSFW images that need structured annotations focused strictly on image-quality issues and safety flags. For each file you will:

• Review the full image, noting any flaws that affect clarity or usability (e.g., blur, compression, artefacts).
• Identify the presence of violence, harmful acts, or other illegal content. When such material appears, label both the exact type of issue and its severity rather than giving a simple yes/no answer.

Nothing else is required—there is no need to tag body positions, clothing details, or general adult-content levels unless it directly supports the violence / harmful-act assessment.

Acceptance criteria
• One annotation entry per image, delivered in JSON or CSV with the image filename, quality notes, issue-type, and severity fields.
• All files processed; spot checks will be performed for consistency and accuracy over a 5% random sample.
• Severity scale remains consistent with the schema provided at project start.

If you have prior experience with large-scale content moderation or computer-vision labelling tools (Labelbox, CVAT, VGG Image Annotator, etc.) this task should be straightforward. I will supply the folder structure, annotation schema, and a short calibration set before you begin.