You bring the idea. OwnClimb brings the climbing vision: turn wall photos into hold data without first becoming a machine-learning team.
OwnClimb handles the computer vision. You get a place to test it, useful geometry, and a small API to build on.
Upload a photo in the playground, inspect the overlay, and tune the confidence threshold before writing integration code.
Every detected hold comes back as a polygon with a confidence score, ready to draw, filter, or turn into your own interaction.
Use the built-in wall and climb tools, or take the same result through the REST API and build the app you had in mind.
Post a wall photo and get scored hold polygons back. Wait for the answer in one request, poll for it, or stream each stage straight into your interface.
curl -X POST https://ownclimb.com/api/v1/detections/blocking \
-H "Authorization: Bearer oc_your_key" \
-F "image=@wall.jpg"
# 200 OK
{
"status": "done",
"image": { "width": 2048, "height": 1536 },
"holds": [
{
"polygon": [412, 233, 430, 251, …],
"confidence": 0.96,
"color": "#c0392b"
}
]
}Start on the personal licence while the idea is yours. Move to Pro when the product is published or making money.
The complete product for experiments, home walls, and the app you are still figuring out.
The same tools and API, with a licence for products you publish, sell, or use in a business.
The practical details behind the demo, API, and plans.
One uncached hold-detection inference from the API or Playground counts as one run. Cached results do not use your allowance; a re-run starts a new inference. Hobby includes 100 runs per month and Pro includes 1,000.
Hobby is for personal projects, experiments, and home walls. Pro adds a commercial-use licence for products you publish, sell, or use in a business; both plans include the same tools and API.
The detector can miss holds or return lower-confidence candidates when a photo is dark, crowded, or badly angled. You can inspect the scored polygons and move the confidence threshold to keep more candidates or fewer, higher-confidence holds; it does not re-run inference.
The API returns scored hold polygons, confidence values and the dominant colour of each hold for a wall photo, along with the image dimensions the coordinates are in. It does not return grades or route suggestions, so your app or route setter stays in control of the problem.
API keys are limited to 120 requests per minute by default, and a 429 response includes Retry-After. Anonymous demo visitors can start up to 10 uncached detections per UTC day.
Yes. OwnClimb stores uploaded images and their derived renditions with your wall and run history; they are not automatically expired after a fixed period. Delete your account in Settings to remove its application data.
Yes. The public demo lets you upload a wall photo and try hold detection before creating an account; anonymous demo use is limited to 10 uncached detections per UTC day.