Skip to content

View inference details

The inference detail page shows the saved configuration for an inference and, when the inference is running, its API connection details.

To open the inference detail page

  1. On the project details page, navigate to the Inference center section.

  2. Click the name of the inference you want to view.

    Project details: Select inference

The inference detail page opens, showing the saved configuration for the selected inference.

Inference info

The inference detail page shows the following inference info.

Section Description
General info Shows the inference name and current status.
API details Available when the inference status is Healthy. Shows the Endpoint URL and a link to the API schema.
  • API endpoint – The API endpoint is where you send the data for predictions. This is also where you'll receive the inference predictions.
  • API schema – Since Robovision AI supports a wide range of algorithms, the data and predictions can come in various formats. For detailed information on the formats, refer to the schema link. This will direct you to a Swagger-generated site with specifics about the API requests and responses.
To copy either value, click Copy link next to the relevant field. Because Robovision AI supports a wide range of algorithms, the data format varies by algorithm; refer to the schema for details.
Info Shows runtime metrics: Throughput, Latency, Requests, and Failures.
Configuration Shows the configuration settings, including the inactive time before shutdown and the number of requested instances.
Model Shows the model used for the inference, its version, and the date it was trained on.
Inference parameters Shows the inference parameters configured for this inference. The available parameters depend on the algorithm.
Area measurement Available for YOLOv8+ instance segmentation and PIDNet semantic segmentation only. Visible only when Detection size measurement is Yes.
Lists all thresholds configured for the inference. Each threshold appears as a collapsible item labeled with the threshold name, expanded by default. Collapse or expand a threshold to see:
  • Classes — one or more classes assigned to this threshold. Hover over it to see the full list if it is truncated.
  • Value — the area condition based on the operator and values configured: is greater than or equal to N px, is less than or equal to N px, or is between N px – M px.
  • Measurement metadata — metadata describing the origin or context of the measurement. Shows None if no metadata was provided. Hover over it to see the full list if it is truncated.
  • Include in sampling — whether this threshold is used to record data during inference: Yes or No.
Sampling parameters Shows the sampling configuration: Random sampling, Confidence-based sampling, and Class-based sampling settings. Also shows the key-value metadata pairs configured for recorded samples.