Explorer
Syncing—

Object Detection in Images - Ultralytics compatible APIService

object-detection

Ultralytics-compatible object detection. The expected endpoints are POST /predict and GET /health; suppliers MAY expose more of the Ultralytics surface but consumers MUST NOT depend on it. This service is model-agnostic: suppliers choose their own detection model and weights, so response quality, class sets, and box coordinates vary and are not guaranteed by the protocol. Every response, including errors, is a single JSON object.

Active Suppliers
5
of 5 ever configured
Service Info
Service ID
object-detection
Compute Units / Relay
35,000
Active Applications
1
Relay Mining EMA
1 relays
Service Metadata Cardsource: on-chain metadata (LCD)
{9 items
"schema":"pocket-service-card/v1"
"description":"Ultralytics-compatible object detection. The expected endpoints are POST /predict and GET /health; suppliers MAY expose more of the Ultralytics surface but cons..."
"rpc_types":[1 item
0:{4 items
"type":"REST"
"intent":"expected"
"backend_hint":"Ultralytics-compatible /predict server on :8080 (image-inference-server); mount at /"
"notes":"Only POST /predict and GET /health are expected. Request body is multipart/form-data or application/x-www-form-urlencoded with file/source, conf, iou, imgsz, no..."
}
]
"apis":[1 item
0:"ultralytics-predict"
]
"specs":[2 items
0:{4 items
"kind":"markdown"
"api":"ultralytics-predict"
"url":"https://raw.githubusercontent.com/pnyxai/pocket-network-services/b687f18c99f71f46316345565aa6e5d0ca02b4a6/mainnet/docs_object-detection.md"
"sha256":"bcdac424babf0c7df43b522e793fd115feef1e81b1a0c4d9e497c237d2d5ad85"
}
1:{4 items
"kind":"openapi"
"api":"ultralytics-predict"
"url":"https://platform.ultralytics.com/openapi.json"
"notes":"Vendor document for the whole Ultralytics Platform, living (no sha256). Request and response shapes of /api/models/{owner}/{project}/{model}/predict apply; this..."
}
]
"access":"public"
"results":"variable"
"serving":{3 items
"backend":"Any Ultralytics-compatible server exposing POST /predict and GET /health. Recommended: image-inference-server (Rust gateway + NVIDIA Triton Inference Server). S..."
"healthcheck":[1 item
0:{4 items
"rpc_type":"REST"
"request":{2 items
"path":"/health"
"method":"GET"
}
"expect":{2 items
"json_path":"$.endpoints.detect.status"
"matches":"^healthy$"
}
"notes":"Pins the detect endpoint: /health reports only enabled endpoints, so an ocr-only deployment fails it."
}
]
"notes":"Use a server with dynamic batching so one backend serves many concurrent requests. Suppliers are responsible for model licensing (e.g. Ultralytics YOLO is AGPL-..."
}
"updated":"2026-10-19"
}