zp.Zakhar Pashkin
← Selected work

Case study

Dermaself Flutter Skin Analysis App

Skin-analysis computer vision for Dermaself, connecting guided mobile capture with pore and wrinkle segmentation and usable results.

Illustrated Dermaself workflow showing facial regions on an AI-generated synthetic face
Illustrated workflow with a synthetic face. The regions are authored illustrations, not model predictions.

Overview

I developed Dermaself's cosmetic skin-analysis computer vision, spanning guided capture, facial regions, pore and wrinkle segmentation, model evaluation and mobile/API integration. The work joined PyTorch and OpenMMLab model development with ONNX and Flutter delivery. I resolved model-asset and runtime differences across cloud and GPU deployments, restoring matching segmentation outputs in regression comparisons. Capture quality, runtime behavior and reproducible evaluation guided candidate release decisions.

Project figures

Illustrated workflow. Model-release evaluation remains separate from an individual image analysis.
Workflow diagram of guided capture, facial regions, selected model runtime and structured results, with separate model-release evaluation
Workflow diagram. Image analysis and model-release evaluation are separate.

What It Covers

  • Guided capture and facial-region processing for consistent model input
  • Pore and wrinkle segmentation with reproducible model evaluation
  • Mobile and API integration across Flutter, ONNX and cloud services
  • Matching regression outputs across cloud and GPU runtimes

Stack And Topics

  • Flutter
  • Dart
  • Firebase
  • Riverpod
  • GoRouter
  • ONNX
  • Mobile CV
  • iOS
  • Android

References

More selected workDownload resumeContact Zakhar ↗