Skin Care Product Recommendation

Category:

AI / FMCG Beauty Care

The Problem Statement

When a customer walks into the exclusive beauty and skin care stores of the client, the store executive recommends a skin care product to the customer after the on-premises detailed skin analysis using an application in a portable device. Most of the time the customer doesn’t trust the suggestions made by the store executive as they are not backed by any data or analysis.

The Solution

To address the convincing of the customer and to make the suggestions more reliable and backed by details, we have developed a skin analysis SDK to analyse the different facial features and skin of the customer using different image processing algorithms, interactive visualizations, and precise recommendations to the same to the customers. This adds the trust and transparency component to the skin care products suggested by the store executive.

Skin analysis features

Blackspot detection

Identifies hyperpigmentation or darkened patches on the skin with pinpoint accuracy. Analyses spot size, type, and pigmentation levels to provide personalized recommendations.

Wrinkle detection

Detects wrinkles or creases that form on the skin because of aging, reduced collagen production, etc.

Yellowness detection

Accurately measures yellowish tone in skin and identifies discoloration levels.

Texture detection

Analyses the overall surface quality of the skin, including smoothness, roughness, and the presence of bumps, pores, or unevenness.

Skin Purity

Detects the level of clarity and cleanliness of the skin, free from impurities like blemishes, acne, blackheads, or uneven tone.

Pore Detection

Detects tiny openings on the skin's surface that release oil and sweat and analyses pore size and condition.

Visualization and analysis

The Visualization module in the skin analysis solution adds interactive and detailed visualization of the current skin condition using customized overlays, which helps consumers to easily understand and view each analysis.

The analysis further combines data and scores from different algorithms to build a collective report, summary and overall score for precise recommendations of products to handle the current underlying problems.

SDK optimization and porting for mobile device

All the algorithms present the Skin Analysis SDK is fully optimized in a native coding language to enable the real-time skin analysis in the low-compute mobile devices and platforms. The SDK can be further optimized and tuned further with respect to target platform like iOS, Android, embedded Linux, etc., and compute platforms, like arm (ARM64 & ARMv7), x86_64 (AMD64 or Intel 64), etc. The minimal tuning and calibration according to camera, condition, and analysis can be performed for enhanced accuracy and performance.

The simplified interface or wrappers developed for different platforms can be utilized to easily integrate and utilize the skin analysis SDK with existing or new applications in any platform (Android & iOS) in minimal time.

Improvements using skin tone classification

Skin tone classification helps us analyse the skin more accurately, this is achieved by detecting the skin tone of each image and setting different algorithm parameters for analysing different skin tones. For example, a black spot might appear differently on a fair skin tone in comparison to a darker skin tone; therefore, some change in the parameters are needed for detecting black spots across different skin tones.

Some of the modules of the skin tone classification algorithm are:

Face detection
Face landmark detection
Extraction of relevant skin pixels
Feature extraction for skin tone classification
Colour classification

Benefits Delivered

Increased customer satisfaction and trust factor with visualization of in-depth skin analysis.
Higher conversion rates as compared to previous financial year.

Gallery

Contacts