https://store-images.s-microsoft.com/image/apps.33738.a404bf91-4b45-462a-b7f8-aafe5d4163e5.b6b7d3f1-5371-458d-9878-f24303659f0b.b87ade0f-4073-440d-a068-fc0b61c754fd

Face Mask Detection API, Maskerizer

SC DE VIS SOFTWARE SRL

Detects and localizes face masks on human faces covering both nose and mouth on a still photo.

Face Mask Detection API - Maskerizer (also known as face mask recognition API) is a cross browsers REST API which get a JSON input with a still photo (as base64 encoded string) or an url of the image and returns a JSON string which contains predictions with certain amount of probability (filtered for output with minimum 50%), bounding boxes of detected face mask(s) put on human faces in a correct way covering both the nose and the mouth with its top, left, width, height values and, if required, for each prediction, a boundingBoxPhoto as base64 encoded string of the detected face mask. The boundingBox contour color may be Green if the detected face mask is put on in a correct way covering both the nose and the mouth, may be Yellow if the face mask put on covers only the nose or only the mouth or may be Red if the person has no mask put on. Also it outputs a base64 encoded string photo which is the original input photo with drawned bounding boxes upon it. For one still photo the API may return multiple predictions with different probability scores of detected face masks put on. Our pricing packages count the predictions, so for one request, multiple predictions may be counted. We filter the results of predictions so we display only the predictions with a probability score higher than 50%. Of course, there are some limitations in order to get a higher accuracy. We recommend properly exposed, unobstructed JPEG photos at 1920x1080 (full HD resolution) where ratio between height of bounding box and height of entire picture should be at least 1:15. For ratios like 1:16, 1:17, 1:18 and so on the accuracy is lower and the AI algorithm may not see the face massk and the prediction score will be low. We do not store pictures. Also, the quality and the angles of the camera are very important and it contribute to a higher detecting accuracy. It should have varifocal lenses, high shutter speed, good infrared lighting beam, full HD resolution.

Allthough this Face Mask Detection API (currently we do not offer a face mask recognition sdk) is intended for software development and therefore developers, we have also here an face mask detection online application that may be used to check the input and output JSONs of the API.

概览
https://store-images.s-microsoft.com/image/apps.1895.a404bf91-4b45-462a-b7f8-aafe5d4163e5.b6b7d3f1-5371-458d-9878-f24303659f0b.7dc80926-2bc5-400e-8552-370b5ecedef3
/images/videoOverlay.png
https://store-images.s-microsoft.com/image/apps.1895.a404bf91-4b45-462a-b7f8-aafe5d4163e5.b6b7d3f1-5371-458d-9878-f24303659f0b.7dc80926-2bc5-400e-8552-370b5ecedef3