detection-job

Check our OpenAPI documentation to check operations and properties of the listed entities: https://api.neurolabs.ai/v1/docs

🔍 Detection Job ~ Task

Images are processed by sending them to a Detection Job, via a REST API call.

When creating a Detection Job, you can have the ability to specify the desired products to detect. Additionally, one can tick the checkboxes to compute a Realogram and Share of Shelf (SoS). The API also allows for listing all available detection jobs, and uploading images to a detection job. The operations can be done both synchronously using /image for single images or asynchronously through /urls for batches of images.

Whilst the v1 API uses Detection Job naming convention for image processing, the equivalent in the web UI is defined as Task and can be found under the Task Management tab.

NeurolabsCoco has some added neurolabs properties on top of the coco json standard format.

{
  "neurolabs": {
    "labels_per_detection": 3
  },
  "info": {
    "year": "2024",
    "version": 1,
    "description": "Coco_Builder",
    "contributor": "Neurolabs Platform",
    "url": "https://www.neurolabs.ai",
    "date_created": "02/20/2024, 11:22:50"
  },
  "processingDuration": null,
  "licenses": [
    {
      "id": 1,
      "url": "https://opensource.org/license/mit/",
      "name": "mit"
    }
  ],
  "images": [
    {
      "id": 1,
      "license": 1,
      "file_name": "https://picsum.photos/1810/935",
      "flickr_url": "",
      "coco_url": "",
      "height": 4032,
      "width": 3024,
      "date_captured": "02/20/2024, 10:22:50"
    }
  ],
  "annotations": [
    {
      "id": 1,
      "image_id": 1,
      "category_id": 2,
      "bbox": [1413, 181, 149, 219],
      "segmentation": [],
      "area": 441,
      "iscrowd": 0,
      "neurolabs": {
        "score": 0.94
      }
    }, 
    {
      "id": 2,
      "image_id": 1,
      "category_id": 1,
      "bbox": [1500, 200, 150, 230],
      "segmentation": [],
      "area": 441,
      "iscrowd": 0,
      "neurolabs": {
        "score": 1, 
        "alternative_predictions": [
              {
                "score": 0.74964449,
                "category_id": 2
              },
              {
                "score": 0.504680041,
                "category_id": 3
              }
        ]
      }
    }
  ],
  "categories": [
    {
      "id": 1,
      "name": "object",
      "supercategory": "",
      "neurolabs": null
    },
    {
      "id": 2,
      "name": "Item 1",
      "supercategory": "",
      "neurolabs": {
        "barcode": "786162004987",
        "customId": "1337",
        "label": "Item 1",
        "productUuid": "00000000-1337-0000-0000-000000000001"
      }
    }
    {
      "id": 3,
      "name": "Item 2",
      "supercategory": "",
      "neurolabs": {
        "barcode": "999162004987",
        "customId": "1338",
        "label": "Item 2",
        "productUuid": "00000000-1338-0000-0000-000000000001"
      }
    }
  ],
  "planogram": null,
  "postprocessingResults": {
    "realogram": [],
    "shares": []
  }
}

COCO Neurolabs Annotation

We have added a score property that is a float and represents the calculated confidence of our inference. 1 representing a 100% confidence. There is also an alternative_predictions property field that highlights predictions below the confidence threshold for which the vision model was deployed. These should be used with caution and further discussed with the Neurolabs Technical team.

COCO Neurolabs Category

In order to map back a generic coco category back to our products, the neurolabs property has some additional information.

{
  "barcode": "1337",
  "customId": "1337",
  "label": "Cola",
  "productUuid": "00000000-1337-0000-0000-000000000001",
  "customId": "your_id"
}