AIRise-ai-fabric-inspection

AIFR-AI – AI-powered fabric defect detection system for Katty Fashion. CNN/YOLO-based real-time quality inspection on Jetson edge devices, integrated with MinIO/NiFi/Spark infrastructure. EU Horizon Europe – AIRISE Open Call 1.

Status

Current Sprint Kanban   Edit Kanban

kanban
  Todo
    t6["⚠️ [BLOCKER] Strategie generare date sintetice — prioritizare GAN/augmentare pentru suplimentare dataset"]
    t7["Design data ingestion pipeline: Apache NiFi + MinIO on-premise"]
    t8["Design data preprocessing pipeline: Apache Spark + augmentation strategy"]
    t9["Define defect taxonomy and annotation schema (Damage, Hole, Knot, Line, Oil Stain, Stain, Wrinkle)"]
    t10["Define CNN/YOLO model architecture for fabric defect detection"]
    t11["Integrate model explainability: Grad-CAM + SHAP pentru vizualizare decizii AI"]
    t12["Backend: FastAPI playback API (Plan B) — endpoints for defect results + annotations"]
    t13["Backend: PostgreSQL/Supabase schema for defect events and annotations"]
    t14["Frontend: PlaybackAnnotator UI — defect review and validation interface"]
    t15["Frontend: TextileViewer component — live WebRTC defect stream view"]
    t16["Jetson setup: YOLO multiprocess detector (4x parallel, 100+ FPS target)"]
    t17["Define S3 storage strategy: MinIO on-premise (active) + Cloudflare R2 (archive)"]
    t18["Define ELK stack monitoring: model prediction logs + feedback loop"]
    t19["Create architecture diagrams and technical documentation"]
    t20["Prepare KPI tracking dashboard (F1 ≥70-80%, inferență ≤1000ms, deșeuri -20%)"]
    t21["Define model retraining pipeline (target: retraining within 48h of new data)"]
    t22["Prepare dissemination content: LinkedIn technical deep-dive post (mid-implementation)"]
  In-Progress
    t3["Define system architecture: edge (Jetson) + cloud (MinIO/NiFi/Spark)"]
    t4["Define streaming architecture — Plan B (Async Playback) selectat ca variantă recomandată"]
    t5["⚠️ [BLOCKER] Audit dataset existent — evaluare volum și calitate imagini defecte reale disponibile"]
  Review
  Done
    t1["Project scope definition & AIRISE initialisation report review"]
    t2["Setup repository structure and KF-CPTO kanban integration"]

Task Summary

LOE Summary

Sprint Timeline

gantt
    title S1 — AIRise-ai-fabric-inspection
    dateFormat YYYY-MM-DD
    excludes weekends

    Project scope definition & AIRISE initialisation report review :done, 2026-03-09, 2026-03-09
    Setup repository structure and KF-CPTO kanban integration :done, 2026-03-09, 2026-03-09
    Define system architecture  edge (Jetson) + cloud (MinIO/NiFi/Spark) :active, 2026-03-09, 2026-03-11
    Define streaming architecture — Plan B (Async Playback) selectat ca variantă recomandată :active, 2026-03-10, 2026-03-10
    ⚠️ [BLOCKER] Audit dataset existent — evaluare volum și calitate imagini defecte reale disponibile :active, 2026-03-11, 2026-03-11
    ⚠️ [BLOCKER] Strategie generare date sintetice — prioritizare GAN/augmentare pentru suplimentare dataset :2026-03-12, 2026-03-13
    Design data ingestion pipeline  Apache NiFi + MinIO on-premise :2026-03-11, 2026-03-12
    Design data preprocessing pipeline  Apache Spark + augmentation strategy :2026-03-13, 2026-03-14
    Define defect taxonomy and annotation schema (Damage, Hole, Knot, Line, Oil Stain, Stain, Wrinkle) :2026-03-13, 2026-03-13
    Define CNN/YOLO model architecture for fabric defect detection :2026-03-13, 2026-03-17
    Integrate model explainability  Grad-CAM + SHAP pentru vizualizare decizii AI :2026-03-17, 2026-03-18
    Backend  FastAPI playback API (Plan B) — endpoints for defect results + annotations :2026-03-16, 2026-03-18
    Backend  PostgreSQL/Supabase schema for defect events and annotations :2026-03-19, 2026-03-20
    Frontend  PlaybackAnnotator UI — defect review and validation interface :2026-03-16, 2026-03-18
    Frontend  TextileViewer component — live WebRTC defect stream view :2026-03-19, 2026-03-20
    Jetson setup  YOLO multiprocess detector (4x parallel, 100+ FPS target) :2026-03-19, 2026-03-20
    Define S3 storage strategy  MinIO on-premise (active) + Cloudflare R2 (archive) :2026-03-20, 2026-03-20
    Define ELK stack monitoring  model prediction logs + feedback loop :2026-03-20, 2026-03-20
    Create architecture diagrams and technical documentation :2026-03-20, 2026-03-20
    Prepare KPI tracking dashboard (F1 ≥70-80%, inferență ≤1000ms, deșeuri -20%) :2026-03-21, 2026-03-21
    Define model retraining pipeline (target  retraining within 48h of new data) :2026-03-23, 2026-03-24
    Prepare dissemination content  LinkedIn technical deep-dive post (mid-implementation) :2026-03-25, 2026-03-25

Effort Distribution

pie title Effort by Status
    "Todo" : 31.0
    "In Progress" : 5.0
    "Done" : 2.0

Links

• Edit Kanban
• Repository
• Kanban Board

Auto-generated by KF Aggregator