EfficientNetB0 MediaPipe Grad-CAM Screening only — not a diagnosis
Select a sample image to analyze

Click any image below — each shows a real behavioral pattern from the dataset. One click → Analyze.

ASD avoid eye contact
avoid eye contact
ASD hand flapping
hand flapping
ASD finger flapping
finger flapping
ADHD continuous moving
continuous moving
Normal normal
normal behavior
Upload your own image

Drop images here or click to browse

JPG, PNG — multiple images supported

How it works
1
Click a sample image or upload your own
2
MediaPipe extracts 8 facial behavioral signals
3
EfficientNetB0 classifies ASD/ADHD symptom patterns
4
Risk scores + Grad-CAM heatmap + clinical report

Analyzing behavioral signals...

01 — Extracting facial signals via MediaPipe

02 — Running EfficientNetB0 classification

03 — Generating Grad-CAM heatmap

04 — Computing ASD & ADHD risk scores

Analysis complete
CNN detected symptom
confidence
ASD risk score
/100
ADHD risk score
/100
Signal breakdown
Grad-CAM heatmap

Red = facial regions that influenced the prediction most

Grad-CAM
Clinical findings


Recommendation