Apppublic

Artz-03/autisme-ai-service

sourceHugging Facemitupdated 2mo agoView on Hugging Face

0likes

App README

AutismeID – AI Service API

Flask backend untuk deteksi indikasi ASD menggunakan MobileViTV2 + FusionBackboneClassifier.

Endpoints

`POST /api/predict`

Upload gambar wajah untuk analisis.

Request: multipart/form-data

—image: file gambar (jpg, jpeg, png, webp, bmp)
—use_tta: "true" | "false" (opsional, default: true)

Response:

json

{
  "prediction": "Autistic" | "Non_Autistic",
  "label": "Indikasi ASD Terdeteksi" | "Tidak Terindikasi ASD",
  "confidence": 0.95,
  "probability_autistic": 0.95,
  "probability_non_autistic": 0.05,
  "risk_level": "high" | "medium" | "low",
  "face_detected": true,
  "cropped_face_b64": "data:image/jpeg;base64,...",
  "use_tta": true,
  "model_version": "mobilevitv2_100_fusion_v1"
}

`GET /api/health`

Cek status service.

Model

—Backbone: MobileViTV2-100
—Source: Artz-03/autismeClassification
—Face Detection: MTCNN (facenet-pytorch)

⚠️ Hasil hanya untuk skrining awal, bukan diagnosis medis resmi.