Skinalor: an Offline, On-Device Dermatology Triage Assistant 📱 Workflow: Photo + description → 2–3 likely conditions → smart follow-up questions → one-tap PDF visit report.

🔥 Project Goal: For many skin concerns, the most time-consuming part of a consultation is “unclear descriptions + missing key questions.” This system organizes essential pre-visit information to save doctor–patient communication time and support triage.

🎬 Full video demo (real mobile on-device inference results):

Upload a skin photo + enter a text description of symptoms

An on-device medical multimodal model runs inference and outputs 2–3 likely condition names (Top predictions)

Based on each predicted condition, the system automatically generates customized follow-up questions (e.g., duration, itching/pain, triggers, changes in location/appearance, etc.)

After the user answers each question, the system summarizes everything: case summary + predicted results + follow-up Q&A

Export a PDF diagnosis/triage report with one tap, ready for in-person visits or online consultations

✨ Highlights:

Bilingual support (Chinese & English)

Offline on-device: minimizes uploads and network dependency, more privacy-friendly

Structured output: turns “fragmented info” into clear bullet points that doctors can scan quickly

Traceable loop: question → answer → summary, making communication more efficient

Built With

Share this project:

Updates