This page details the endpoint for sending images for classification through Foyer Insight. An SDK is provided at https://www.npmjs.com/package/@foyer-inc/insight-sdk
There is a 10MB limit on images sent to the classify endpoint.
Classify a single image
POSThttps://api.foyer.ai/Media/classify
This endpoint runs your image through classification and object detection. Accepts a single image with the parameters below, or an array of images.
Attributes are more specific features gathered from the objects detected in the images. For the segmentations that have them, the attributes will give more robust information about the image. Attributes will always have a name field and a value field. Currently, the possible attributes are:
tagpoint: the tagpoint attribute's value is an array in the shape of [x, y] where x and y are percentage-based. They form a point that is guaranteed to be on the segmentation for tagging/display purposes.
is_stainless: A boolean. This attribute appears on refrigerator, stove, oven, washing_machine, hood, and dishwasher segmentations.
floor_type: A string of one of the possible flooring classifications (defined below). An additional float field, confidence, exists on this attribute.
Possible flooring classifications
A floor segmentation will automatically have a flooring type within its attributes.
Classify images from a RESO object
POSThttps://api.foyer.ai/Media/classifyReso
This endpoint runs multiple images through classification and object detection from a valid RESO payload.
This function accepts a RESO payload and returns inference data from all provided Media objects. This returns the RESO data with additional classification and segmentation data added to it. To fit RESO guidelines where possible, output classifications are contained in the ImageOf field.
The classifyRESO endpoint does not return mask data. Mask data for images on a full listing can make the resulting payload many times larger. We can provide bulk inference with mask data included, but logistics for the large payloads must first be arranged.