Use Case Title
Barcode Detection and Recognition System
Team Members
Ruchita Patil
Sanskruti Patil
Sarvasvi Patil
Gauri Potdar
**Use Case Importance **
Barcode detection is widely used in retail, inventory management, logistics, and product tracking systems. Our solution helps automate barcode identification in real time, reducing manual work and improving efficiency and accuracy.
Data Collection and Annotation
We collected barcode images from different products under various lighting conditions and camera angles using mobile phones. The dataset included barcodes of different sizes and orientations to improve model performance in real-world situations.
Annotated Classes:
- Barcode
Annotation Tool Used:
- Roboflow / LabelImg / Makesense.ai
Total Images:
-
1175 Images
- Training Images: 826
- Validation Images: 236
We carefully annotated the barcode regions to improve detection accuracy during training.
Model Training and Validation
We trained the model using YOLOv9t for barcode detection and monitored metrics such as mAP, precision, recall, and loss values during training and validation.
Training Details:
- Model Version: YOLOv9t
- Training Images: 826
- Validation Images: 236
- Detection Accuracy: 98.4%
After initial testing, we included more diverse barcode samples to improve the model’s ability to detect barcodes under different lighting and viewing conditions.
Model Deployment and Demo Video
The trained model was successfully deployed on the YOLOv9 mobile application for real-time barcode detection. The system performed efficiently during live testing and was able to detect barcodes accurately in different environments.
Deployment Details:
- Number of Models Deployed: 1
- Approximate FPS: Yet to be measured
- Detection Accuracy: 98.4%
Demo Video Link:
[(https://drive.google.com/file/d/1MspdaDtL_bNcPqAhraVBGTIENXzOXyOr/view?usp=drivesdk)]
Conclusion
Our barcode detection system achieved high accuracy in real-time testing and demonstrated the practical use of computer vision in automation tasks. Through this project, we learned the importance of high-quality datasets, proper annotation, and diverse training samples for improving model performance and reliability.