Acne Grading API

[General Instructions] | [Demo]

Acne is the commonest ‘disease’, but severe forms are associated with considerable morbidity. Dermatologists grade acne according to severity to decide when to intervene and treat it. As acne is extremely common, computer-assisted acne grading is a popular machine-learning task.

So far the popular approach has been to create CNN-based classification models. It is difficult to train and deploy these models and the clinical utility is limited. There have also been regression-based approaches as the grades are ‘ranks’ than ‘classes’. We adopt an object detection-based approach that categorizes acne lesions into four basic types: comedones, papulopustular, nodulocystic and scars.

After characterizing and counting these subtypes, we use a weighted counting to reduce it to a single numeric grade between 0-8. The API returns the submitted image, along with the lesion counts and this single numeric grade.

This API is suitable for cosmetic dermatology service providers to document progress and to screen clients to identify severe cases. The detection sensitivity can be adjusted according to the needs.

Connect on RapidAPI


The scar detection is suboptimal. We will try to improve this in future versions. The backend improvements may not require any changes in the frontend.

Image requirements:

Submit a close-up image of the side of the face with neutral lighting. Try it out few times to find an appropriate sensitivity for your use case and image acquisition system.


Output includes the image with the detected lesions boxed. Additionally, the counts of the four lesion subclasses and an overall numeric grade are returned.

Sample output

{"function": "AcneGrading", "image": "data:image/jpeg;base64, /9j/4AAQSkZJRgABAQEAZABkAAD/2wBDAAIBAQEBAQIBAQECAgICAgQDAgICAgUEBAMEBgUGBgYFBgYGBwkIBgcJBwYGCAsICQoKCgoKBggLDAsKDAkKCgr/2wBDAQICAgICAgUD9k=", "comedone": 17, "papulopustular": 9, "nodulocystic": 0, "scar": 0, "acne_grade": 0}
API Demo
Click here for a free demo!

Skin Tone API

[General Instructions | Demo]

Skin tone, the genetically endowed amount of melanin in the outer skin layer, has got enormous significance both culturally and cosmetically. It is interesting to note that some consider less is better while others strive for more. Fitzpatrick’s skin type, though primarily not designed for grading skin tone, is the widely accepted gold standard.

Measuring the amount of melanin is an arduous task and estimates from digital image analysis are extremely error-prone. The estimate may vary significantly dependent on lighting and image quality.

This API estimates approximate tone by a pixel counting method after converting the submitted image to an appropriate colour space. The estimate is approximate and this is useful only for cosmetic purposes.

Connect on RapidAPI

Image requirements:

Submit an image with only (or predominantly) skin pixels. Take a close-up image with neutral lighting. Try it out few times to get it right.


Output includes the skin tone in ITA. Read more about ITA here. In addition to skin tone, the API also provides an approximate estimate of erythema (redness) and cyanosis (blueness). Please note that both are NOT appropriate for clinical purposes.

Response: In the ‘Result‘ array, look for the ‘Title‘, ‘Detail‘, and the ‘ValueCompute‘ fields.

API Demo
Click here for a free demo!

Open-Source Flutter App

We are not professional app builders or chatbot designers. We provide backend services that make them intelligent. This is a prototype Flutter app to demonstrate the integration with our backend services offered through RapidAPI. This is open-source and is available on GitHub to modify and use.

Take a photo under natural lighting. You can switch cameras.

Zoom to the skin!

Crop and view the results. (Sends image to our server, but the image is discarded after analysis)

The full source code of this prototype is available on GitHub!