YOLOv5 Industrial LED Defect Detection

Problem Statement

This project implements an advanced YOLOv5-based system for industrial LED quality control using Object-Oriented Bounding Boxes (OBB) to precisely detect and classify LED defects. The system is specifically designed to identify:

🔍
Missing LEDs

Detection of absent LED components in expected positions

🔄
Misoriented LEDs

Identification of LEDs with incorrect rotational alignment

Performance Results

Comprehensive performance evaluation of our optimized YOLOv5 models across different deployment scenarios, demonstrating significant improvements in both accuracy and efficiency.

Performance

Model
mAP
FPS
GFLOPs
Baseline
96.7
43.05
17.7
Version 1
98.2
32.32
64.2
Version 2
98.2
42.64
21.3

Note: FPS is calculated during model inference only, without considering the preprocessing and Non-Maximum Suppression (NMS) steps.

Sample Detection Results

Examples of our Object-Oriented Bounding Box (OBB) detection system in action, showcasing precise identification of missing and misoriented LEDs in industrial settings.

Sample Output 1

Missing LED Detection
Sample detection output 1
🖼️
Sample Output Image 1
Please update image path

Demonstration of missing LED detection using Object-Oriented Bounding Boxes with precise localization and classification.

Sample Output 2

Misoriented LED Detection
Sample detection output 2
🖼️
Sample Output Image 2
Please update image path

Example of misoriented LED detection showing the system's capability to identify rotational alignment issues with high accuracy.