Releasedv1.0.0Computer Vision · Deep Learning

Computer Vision Inspection Platform

Multi-task computer vision inspection platform: classification, detection, segmentation, XAI, and FastAPI serving.

Multi-task industrial inspection: classification, segmentation, segment-then-box detection, Grad-CAM, FastAPI.

Interactive Demo GitHub Metrics Architecture API Docs Presentation README

Problem

Quality inspection needs more than a notebook

YOLO demos alone do not prove platform skill

Multi-task CV needs shared serving contracts

Baselines and honesty matter for trust

Solution

A CV inspection platform, not a single model

Classify · detect · segment · explain
Baselines vs primary with real metrics
FastAPI multipart inference

Pipeline

Interactive system flow

Selected stage

Synthetic panel

Industrial-style inspection imagery.

Architecture

System architecture

Download

Interactive Demo

Use the real system

Start with Run Example — one click fills a realistic payload and calls the live FastAPI via this site's proxy.

Capability focus: multi-task inspection — classify with Grad-CAM, detect with boxes, segment with mask overlay.

Run Example loads sample img_0012.png and shows Grad-CAM on the source image.

Original

Upload or Run Example

Visual result

Classify → Grad-CAM · Detect → boxes · Segment → mask

// CV response JSON

Loading Interactive Demo…

Performance

Measured evaluation results

Numbers below match this project's evaluation reports. Dataset scope, baselines, and limitations are documented under Integrity.

Primary F1

1.000

Baseline 0.933

Det mAP@0.5

0.958

Baseline 0.958

Seg mIoU

0.993

Comparison chart · source: project reports/metrics

Metric	Primary	Baseline
Primary F1	1.000	0.933
Det mAP@0.5	0.958	0.958
Seg mIoU	0.993	—

Loading Performance…

Engineering Quality

Production packaging signals

PyTorch

Multi-task models

Grad-CAM

Explainability

Docker

Dockerfile

GitHub Actions

MLflow

Metrics logging

Docs

Engineering docs 00–28

Integrity

Engineering decisions & limitations

Synthetic industrial sample — not a live plant warranty
Baselines reported for classification and detection
Detection primary uses documented segment-then-box
Compact CPU-friendly models for reproducible demos
Real metrics only after training runs

Repository

Source of truth

README Architecture Presentation Dashboard Repository Release v1.0.0

Presentation

Hiring presentation pack

Presentation pack Performance dashboard Architecture page

API Docs

Execute live FastAPI contracts

// Response JSON appears here

Loading API Explorer…

README

Source documentation

Open project README →

About Engineer

Mohammad Ahmadian

AI / Machine Learning Engineer

Production-oriented AI/ML systems — evaluation, APIs, and honest limitations. Turkey (GMT+3).

GitHub Email Resume Contact