This endpoint analyzes an image for fake faces. Unlike requests made with an image_url, all file objects are ‘interpreted’ with a color heatmap.

For authorization, the request must include an x-api-key header with a value of {api-key}:{api-secret}, for example: KLJId2323sdcvd3:3ASDsdf2323s_df

The request accepts a form-data body with a file key mapped to a value containing the image file to be analyzed.

The endpoint returns a JSON object with keys:

• fake_prob, a float between 0 and 1 of the model’s estimation of the probability the image is GAN-generated.

• facebox, a list of float coordinates representing the corners of a bounding box of the face analyzed by the system.

• face_confidence, a float between 0 and 1 that represents the model’s confidence that a face, real or synthetic, is in the image. Images with low face confidence should be rejected.

• input_size, a coordinate representing the size of the input image.

• interpreted_image, a string of bytes representing the PNG-encoded orange/blue color heatmap of the image representing the areas of the image that look ‘more’ or ‘less’ like a synthetic image to the model.

This endpoint analyzes an image for fake faces and optionally returns an interpretable result.

For authorization, the request must include an x-api-key header with a value of {api-key}:{api-secret}, for example: KLJId2323sdcvd3:3ASDsdf2323s_df

The request body accepts a JSON object with keys:

• url, a string representing the url of the image to analyze

• interpretable_output, a boolean that indicates whether the API should return an orange/blue color heatmap of the image representing the areas of the image that look ‘more’ or ‘less’ like a synthetic image to the model. Note that this is only available for images submitted by url, not images submitted by a file object.

The endpoint returns a JSON object with keys:

• fake_prob, a float between 0 and 1 of the model’s estimation of the probability the image is GAN-generated.

• facebox, a list of float coordinates representing the corners of a bounding box of the face analyzed by the system.

• face_confidence, a float between 0 and 1 that represents the model’s confidence that a face, real or synthetic, is in the image. Images with low face confidence should be rejected.

• input_size, a coordinate representing the size of the input image.

• If interpretable_output is true, interpreted_image, a string of bytes representing the PNG-encoded orange/blue color heatmap of the image representing the areas of the image that look ‘more’ or ‘less’ like a synthetic image to the model.

This endpoint displays the month-to-date number of requests made with this API key. For example, if it is 12:32:01 July 7th UTC, it will display all requests made since 00:00:00 July 1st UTC. Requests made to this endpoint count toward the total.

For authorization, the request must include an x-api-key header with a value of {api-key}:{api-secret}, for example: KLJId2323sdcvd3:3ASDsdf2323s_df