Skip to content

ProjectsComputer Vision Inspectionreadme

README

Source documentation

Computer Vision Inspection Platform

Status Python License

Mohammad Ahmadian — AI / Machine Learning Engineer
Target roles: Computer Vision Engineer · Machine Learning Engineer

Enterprise-Grade Visual Inspection System for Industrial Quality Control

Hiring managers: Demo Website · Dashboard · Architecture · Presentation pack

Positioning

This is a Computer Vision Inspection Platform, not a single YOLO notebook.

Capabilities (v1):

  • Image classification (OK / defect) with Grad-CAM
  • Object detection via segment-then-box (+ classical baseline)
  • Semantic segmentation (Compact U-Net)
  • FastAPI serving, baselines, evaluation, Docker, tests

Recruiter snapshot

| Item | Detail | |------|--------| | Domain | Industrial CV · Visual inspection · Quality control | | Tasks | Classification · Detection · Segmentation · XAI | | Serving | /health · /v1/classify · /v1/detect · /v1/segment · /v1/model | | Integrity | Baselines vs primary · real metrics · synthetic sample honesty |

Evaluation (synthetic industrial sample · test)

| Metric | Value | |--------|------:| | Primary classification F1 (ResNet18) | 1.000 | | Baseline classification F1 (LogReg) | 0.933 | | Detection mAP@0.5 proxy (segment→box) | 0.958 | | Segmentation mean IoU | 0.993 |

Full report: reports/EVALUATION_REPORT.md

Architecture

System architecture (Released)

Quick start

python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
pip install -e .

set PYTHONPATH=src
python scripts/prepare_dataset.py
python scripts/train_and_eval.py
uvicorn cvinspect.api.main:app --app-dir src --host 0.0.0.0 --port 8000
# http://localhost:8000/docs

Docker:

docker compose up --build

Documentation

docs/00_MASTER_INDEX.md · Doc 13: Computer Vision Pipeline

Limitations (honest)

  • Seeded synthetic industrial-style imagery — methodology demo, not a live plant warranty
  • Compact CPU-friendly models for reproducible portfolio packaging
  • Detection primary uses segment-then-box (documented engineering choice)

Contact

mohammad.ahmadian.dev@gmail.com · github.com/ahmadian-dev · Turkey (GMT+3)

License

MIT — see LICENSE.