TL;DR
Get beauty and skincare favorites delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
Five master’s students from Politecnico di Milano and Politecnico di Torino developed image-processing technology with Dermaself for AI skin analysis. The system is designed to standardise selfie conditions and reduce facial identifiers while preserving visible skin details; independent performance and privacy validation were not provided in the report.
Five Alta Scuola Politecnica master’s students have developed image-processing tools with beauty technology company Dermaself to make selfie-based AI skin analysis more consistent and less identifiable. The project combines guidance on photo capture, image correction and a 3D facial reconstruction method intended to reduce biometric identifiers while retaining visible skin features needed for analysis.
The work targets two problems with using consumer selfies as inputs to skin-analysis software. Smartphone cameras can alter the appearance of skin through automatic exposure, white balance and image processing, while differences in lighting and distance create further variation. Those changes may affect how consistently an AI system detects visible characteristics such as redness, spots, lesions and texture. The project’s first component guides users on distance and lighting before capture, then applies a hybrid processing pipeline that combines AI-driven illumination correction with conventional image-processing methods.
The second component addresses the fact that a face image can contain both skin information and identifying facial features. Rather than relying only on blur, masking or pixelation, the team developed a 3D reconstruction and anonymisation pipeline. According to the project description, it modifies the reconstructed facial geometry to reduce biometric identifiers and then re-projects the original high-resolution skin texture onto that altered structure. The aim is to retain visible skin detail while making the person less recognisable.
The five students came from two universities: Andrea Germano, Adriano Giuliani and Federico Greppi represented Politecnico di Torino, while Edoardo Gribaldo and Alessia Soccionovo represented Politecnico di Milano. Academic tutors were Professor Elisabetta Raguseo, director of Alta Scuola Politecnica, and Professor Federica Arrigoni. The project was undertaken in direct collaboration with Dermaself, whose platform analyses visible skin characteristics from selfies and connects results to skincare products in a brand’s or retailer’s catalogue.
Making Skin Analysis More Consistent
The development matters because selfie-based analysis depends on images captured in conditions the software does not control. A customer might use a phone at home, in a shop, at a brand event or through an e-commerce page. More consistent image inputs could help an analysis system compare visible characteristics across those settings, although the source does not report measured gains in accuracy or reliability.
The privacy method also addresses a practical tension for companies and researchers: facial anonymisation can remove the same details that skin-analysis tools are designed to inspect. If the approach works as intended, it could let a platform retain useful skin texture while reducing identifying information in images. That may be relevant to consumer trust and the handling of face images, but the report does not establish that the method makes images fully anonymous or satisfies any particular legal standard.
For retailers and beauty brands, Dermaself’s system is designed to connect image analysis with product recommendations from their own catalogues. The student project therefore sits between academic image-processing research and a commercial customer experience. Its importance will depend on technical validation and how the tools are incorporated into real services, not just on the stated design goals.
As an affiliate, we earn on qualifying purchases.
From Camera Variability to Privacy
Images taken on different phones or under different lighting can look unlike one another even when they show the same person and skin. Camera settings may brighten, darken or alter colour balance automatically. For software intended to identify subtle visible characteristics, these differences are a source of variation. The Alta Scuola project’s capture guidance and post-capture correction are intended to reduce that variation before Dermaself’s analysis runs.
Conventional methods of obscuring a face can lower its recognisability, but may also hide areas of skin that an analysis needs. The project’s alternative separates the facial geometry from the texture in its processing pipeline: it changes the reconstructed structure and reapplies the skin texture. This is the reported design approach; the source does not provide comparative testing against blurring or other anonymisation methods.
Alta Scuola Politecnica is the academic programme through which students from Politecnico di Milano and Politecnico di Torino worked on the project. Dermaself’s platform is intended for use across online retail, physical beauty settings and brand events. The collaboration was framed as an effort to move student research toward use in a functioning AI skin-analysis ecosystem.
Testing and Privacy Results Pending
The report describes the system’s design and intended uses but provides no accuracy measurements, test results or comparison data showing how much image consistency improves. It also does not say how reliably the altered images resist identification, whether the method has undergone independent privacy review, or what happens to original selfies and processed images after analysis.
It is not clear whether the tools are already deployed in Dermaself’s products, being tested with users, or remain at a research and development stage. The source also does not detail the size or makeup of any evaluation dataset, the range of devices tested, or how the system performs across different skin tones and capture conditions. Those details are needed to judge both practical performance and the limits of the approach.
Deployment and Validation Remain Open
The next reported milestone is not specified. The source does not give a launch date, a testing schedule or a plan for independent evaluation. Further information from Dermaself or the academic team could clarify whether the tools will be integrated into the company’s customer-facing services and how their performance will be measured.
For the project’s claims to be assessed, useful next disclosures would include results from testing across cameras, lighting conditions and skin characteristics, alongside details of privacy safeguards and image retention. Until such information is available, the methods should be understood as technologies designed to improve consistency and reduce identifiability, rather than as proven guarantees of accuracy or anonymity.
Key Questions
What did the Alta Scuola Politecnica students develop?
They developed two image-processing components for Dermaself: a pipeline intended to standardise selfie lighting and capture conditions, and a 3D facial reconstruction method designed to reduce identifying features while retaining visible skin texture.
How is the project intended to protect identity?
The described method alters the geometry of a reconstructed face and re-projects the original skin texture onto it. The stated goal is to reduce biometric identifiers without discarding visible skin details. The report does not provide independent evidence that the output is fully anonymous.
What does Dermaself do with a selfie?
Dermaself’s platform analyses visible skin characteristics in a selfie and uses the results to generate skincare recommendations from a brand’s or retailer’s product catalogue.
Has the technology been proven to improve analysis accuracy?
The source describes the intended benefits but gives no measured accuracy results or independent validation. The extent of any improvement remains unclear.
When will the tools be available to consumers?
The report does not state whether the components are already deployed or give a public release date. Their development and integration status has not been specified.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
