Hive’s main logo model is the world’s most comprehensive logo model, boasting tens of millions of training labels and capable of recognizing 5000+ logos. Please contact our sales team for the complete list or to inquire whether a specific logo exists.

Hive’s logo model response format is an instantiation of the detection response format that outputs logos and logo placements — also known as locations. If specified, a model can also only return logos without the associated logo placements.

Logos and locations are returned as separate objects in the bounding_poly list. Additionally, if a logo is identified to be placed on a location, the location will be specified in that logo detection's meta field.

{
  "output": [
    {
      "time": 6.973633333333334,
      "bounding_poly": [
        {
          "dimensions": {
            "top": 478.2709655761719,
            "bottom": 524.2686767578125,
            "left": 191.23072814941406,
            "right": 294.96636962890625
          },
          "vertices": [
            {
              "x": 191.23072814941406,
              "y": 478.2709655761719
            },
            {
              "x": 294.96636962890625,
              "y": 478.2709655761719
            },
            {
              "x": 294.96636962890625,
              "y": 524.2686767578125
            },
            {
              "x": 191.23072814941406,
              "y": 524.2686767578125
            }
          ],
          "classes": [
            {
              "class": "krispy_kreme",
              "label": "Krispy Kreme"
            }
          ],
          "meta": {
            "id": "32650d10-7db4-11ea-8655-53dce52128d7",
            "type": "logo",
            "clarity": 0.001390202034425103,
            "locations": [
              {
                "id": "327560c0-7db4-11ea-8655-53dce52128d7",
                "label": "car",
                "score": 0.9988117218017578
              }
            ]
          }
        }
      ],
      "subtask_no": 269094
    }
  ]
}

The additional fields to the detection response format that are unique to the logo models are:

Name

Description

classes.0.class

Sanitized representation of the logo or location that is lower-cased and free of special characters.

classes.0.label

A “prettified” representation of the class for the brand name or location, including special characters and capitalization.

meta.type

The type of detection: logo or location.

meta.clarity

A score for the clarity of the detection, with 1 being the most clear and 0 being the least clear.

meta.locations

Only applies to logo detections. Contains the location (if any) associated with the logo detection, as well as its label and confidence score respectively.

Given the large number of logos and the variance of the confidence score across different logos, the confidence score for each detection is not returned in the logo model response. Instead, each logo class is carefully thresholded by Hive, such that only logos above a certain degree of certainty are returned to the user.