Summary
Counting things in a photograph stops being a manual job.
A text prompt narrows detection to the object you care about, and the answer comes back with counts and 2D coordinates rather than a description of the scene. visualize-detection-result writes the boxed image to disk, which is what you hand to somebody who does not believe the number.
What it is
Object detection over an image URI in four tools: prompt-driven detection, full-scene detection, human pose keypoints, and a visualiser that draws boxes and labels onto the original image.
What you get
- Objects matching a text prompt identified and counted, with their 2D coordinates and, on request, descriptions
- Every identifiable object detected without a prompt, returning category, count, coordinates and descriptions
- 17 keypoints per person in an image, for body posture and movement analysis
- Detections drawn onto the original image with box thickness, font size and labels under your control, saved to the configured image directory
Requirements
A DINOX_API_KEY and an IMAGE_STORAGE_DIRECTORY for the annotated images to be written into, plus npx on your PATH.
Setup effort
One command plus a key — npx -y @deepdataspace/dinox-mcp, then supply credentials