// IMAGE & VISION
Image Moderation and Watermark Detection
Two nodes screen incoming images. IMAGE__NSFW_GATE detects explicit content and stops the pipeline when it finds some. IMAGE__WATERMARK finds watermarks and returns where they are and how confident the detection is.
Nodes used
IMAGE__NSFW_GATENSFW detection and filter. Blocks the pipeline if explicit content is detected.
IMAGE__WATERMARKDetects and localizes watermarks. Returns bounding boxes and confidence.
How to set it up
- 01Place IMAGE__NSFW_GATE directly after INPUT__IMAGE so blocked files never reach later nodes.
- 02Add IMAGE__WATERMARK if you need to know whether and where a watermark appears.
- 03Publish and call the endpoint. A blocked image ends the run instead of returning processed output.
Use it as a gate, not a verdict
Classifiers make mistakes in both directions. Treat the gate as a first filter and keep a human review path for borderline or high-stakes content.
Order matters
Putting moderation first saves GPU time on rejected uploads. Watermark boxes are a detection result only; they do not remove the watermark.