Capstone Project · Tarlac State University · 2026

Your personal
skin companion.

DermaLens uses an on-device YOLOv11 model to detect six skin conditions from a photo — track progress and find clinics without leaving the app.

Download APK GitHub

6Conditions Detected

65.4%Overall mAP50

100%On-Device AI

Recent Scan

Eczema 67.4%

2 days ago

Quick Actions

Scan Skin

Find Clinics

Progress

Diagnostic aid only. Consult a dermatologist.

Privacy-First• Real Bounding Boxes• Progress Tracking• Clinic Locator• Anonymous Research• Tarlac State University• AI-Powered Detection• On-Device Inference• YOLOv11 Model• 6 Conditions• Privacy-First• Real Bounding Boxes• Progress Tracking• Clinic Locator• Anonymous Research• Tarlac State University• AI-Powered Detection• On-Device Inference• YOLOv11 Model• 6 Conditions•

DermaLens is a diagnostic aid only. Always consult a licensed dermatologist for professional advice.

What it does

Everything you need to understand your skin.

AI Skin Scan

Point your camera at any skin concern. YOLOv11 runs entirely on-device, drawing real bounding boxes with condition, confidence, and severity.

Progress Tracker

Trends per tracked spot, so two unrelated warts never merge into one timeline. Every scan keeps its photo and editable notes — tap any entry to revisit the full result.

Clinic Locator

GPS-powered map finds nearby dermatology clinics via Google Places, with driving routes and ETAs. Users can confirm whether a clinic is still open, and permanently closed places are filtered out.

Research Contribution

After saving a scan, opt in to share it anonymously over Wi-Fi. You choose per scan — it is never a silent side effect of saving.

Accessibility

Font size slider and high contrast mode propagate across every screen — manually verified across all conditions and UI states.

Process

Three steps to a result.

Capture

Use your camera with pinch-to-zoom, or pick from gallery. Position the affected area inside the guide frame.

Analyze

YOLOv11 runs on-device in seconds. Bounding boxes appear with condition name, confidence score, and severity.

Act

Save to Progress, find a nearby clinic, or contribute anonymously to the research dataset.

Demo

See it in action.

AI Model

YOLOv11, trained on six conditions.

The v2 model is a merged 6-class YOLOv11-small, exported to TFLite and verified live on-device. Instance-aware oversampling eliminated the v1 model's cross-condition confusion — the off-diagonal ceiling is now just 0.09.

65.4%

Overall mAP50

30%

Confidence threshold

0.09

Off-diagonal max

On-Device

Runtime

Detection Accuracy per Condition

Confidence threshold: 30%. Results below this floor show a "Low Confidence" prompt instead of a shaky guess.

Tip of the day — Don't over-zoom. Cropping tightly on a lesion measurably lowers confidence; frame it with surrounding skin, like the training photos do.

The Team

Built by four students at Tarlac State University.

Mark Joseph Garcia

AI Model Integration

Reynaldo Manio Jr.

Camera & YOLOv11 Training

Reicee Owen Pastrana

Scan Result

Chrisent Dayniel Tolentino

Auth & Progress Tracker

Android · API 26+

Take it for a scan.

DermaLens is a student research project. All inference runs on your device — no images leave unless you explicitly choose to contribute.

Download APK GitHub