Likeness Detection

Overview

The Likeness Detection model classifies an entire image based on the presence of the most popular characters. It's powered by a classification model, with each class representing one character and returning a confidence score between 0 and 1.

The model does not only detect exact representations of IP, but also drawings, fanart, costumes, and AI generated versions of the content. This model is also resistant to image manipulations like rotations and text overlays, as well as more subtle augmentations such as introduction of noise, filters, and other pixel-level changes.

Request Format

# submit a task with media with url
curl --request POST \
  --url https://api.thehive.ai/api/v2/task/sync \
  --header 'accept: application/json' \
  --header 'authorization: token <API_KEY>' \
  --form 'url=http://hive-public.s3.amazonaws.com/demo_request/faces.jpg'

# submit a task with media with local media file
 curl --request POST \
     --url https://api.thehive.ai/api/v2/task/sync \
     --header 'Authorization: Token <token>' \
     --form 'media=@"<absolute/path/to/file>"'

Response

The output of our Likeness Detection model is formatted as a list of classes, where each object in the list contains the following attributes:

class: the name of a likeness detected in the image or frame
score: the confidence score assigned to the detected class
parent_company: the company that owns that likeness
category: the likeness type detected, which will be one of entertainment_media (movie, TV show, or video game character), company_mascot, toy, or other

To see an annotated example of an API response object for this model, you can visit our API reference page.

Thresholds

We recommend a threshold of 0.9 for optimized model performance.

We recommend that you start off with these thresholds, but you should always refine these thresholds to suit your specific use case.