End-to-End System for iPhone Backside Panel Inspection

Defect detection · Size measurement · Severity grading (A–D)

Overview

Complete inspection pipeline that detects defects, measures their size, and grades severity from A to D. Multi-stage system combining detection, measurement, and classification for automated quality control.

🎯 Detection 📏 Measurement 📊 Grading Python OpenCV PyTorch

Problem

  • Need for automated severity classification beyond binary OK/NG.
  • Defect size measurement required for grading criteria.
  • Varying surface appearance affects detection consistency.
  • Multiple defect types with different severity thresholds.

Solution

  • Multi-stage pipeline: detection → measurement → grading.
  • Severity classification (A–D) based on size and criteria.
  • Robust feature extraction for varying surface appearance.
  • Automated decision support for quality control.

My Role

  • Designed multi-stage inspection with detection + measurement.
  • Implemented severity classification based on size/criteria.
  • Improved robustness for varying surface appearance.
  • Validated grading accuracy against human inspection standards.

Implementation

Language: Python

Libraries:

OpenCV PyTorch

Severity Grades: A (minor) → D (severe)

Results Gallery

Training and evaluation demonstrations are omitted due to data confidentiality.

End Result

  • Accurate defect grading and automated decision support.
  • Consistent severity classification across production batches.
  • Reduced manual inspection time and improved QA consistency.