AI-Powered PCB Defect Detection

Automated quality control for printed circuit boards

Python TensorFlow OpenCV Computer Vision
GitHub Live Demo
AI-Powered PCB Defect Detection

Project Overview

This project implements a real-time defect detection system for printed circuit boards (PCBs) using computer vision and deep learning techniques. The system can identify various types of defects including soldering issues, component misalignment, and missing parts.

Key Features

  • Real-time Detection: Process images at 30+ FPS
  • Multi-class Classification: Detect 15+ different defect types
  • High Accuracy: 98.5% accuracy on test dataset
  • Easy Integration: RESTful API for seamless integration

Technology Stack

  • Deep Learning: TensorFlow 2.x with Keras
  • Computer Vision: OpenCV for image processing
  • Detection Architecture: YOLOv8 for object detection
  • Backend: Flask API server
  • Frontend: React dashboard for monitoring

Results

The model was trained on 10,000+ labeled PCB images and achieved:

  • Precision: 98.2%
  • Recall: 97.8%
  • F1 Score: 98.0%
  • Inference Time: 33ms per image

Implementation

1
2
3
4
5
6
7
8
9
10
11
12

import cv2
import tensorflow as tf

# Load the trained model
model = tf.keras.models.load_model('pcb_detector.h5')

# Process image
def detect_defects(image_path):
    image = cv2.imread(image_path)
    preprocessed = preprocess_image(image)
    predictions = model.predict(preprocessed)
    return parse_predictions(predictions)

Challenges & Solutions

  1. Class Imbalance: Used data augmentation and weighted loss functions
  2. Real-time Processing: Optimized model with TensorRT
  3. Lighting Variations: Implemented adaptive preprocessing

Future Improvements

  • Deploy to edge devices (NVIDIA Jetson)
  • Add 3D defect detection capabilities
  • Implement active learning pipeline
  • Expand to other PCB types

Conclusion

This project demonstrates the power of AI in automating quality control processes, significantly reducing inspection time while improving accuracy.