Price Tag Detection using YOLO Object Detection
Team Members
- Umbare Shivprasad
- Yadav Nehru
- Patil Krish
- Patil Shreyash
Use Case Importance
Price tag detection systems are highly useful in smart retail stores, supermarkets, inventory management, automated billing systems, and digital shelf monitoring. Detecting price tags in real time helps businesses manage products efficiently and reduce manual work.
Our project uses Computer Vision and YOLO-based object detection to detect and locate price tags from product shelf images with fast and efficient real-time prediction.
Data Collection and Annotation
We collected and prepared a custom dataset consisting of multiple product shelf images containing price tags captured from online sources and local markets under different lighting and viewing conditions.
The dataset included:
- Supermarket shelf images
- Grocery store products
- Multiple price tag positions
- Different lighting conditions
- Various camera angles and distances
Images were annotated and converted into YOLO format for training and detection purposes.
Classes used:
- “Price Tag”
The dataset was preprocessed and organized into:
- Training Dataset
- Validation Dataset
- Testing Dataset
Preprocessing techniques included:
- Image resizing
- Data normalization
- Data augmentation
- File organization
- YOLO format conversion
These preprocessing steps improved training efficiency and model performance.
Sample image:
Model Training and Validation
We trained the YOLOvX object detection model using the prepared dataset containing product shelf images with price tags.
Training parameters such as:
- Epochs
- Batch size
- Learning rate
were adjusted to improve model accuracy and stability.
Metrics observed during training included:
- Detection accuracy
- Prediction confidence
- Model loss
- Precision and recall performance
The model successfully learned price tag features and detection patterns.
During testing, some inaccuracies were observed in blurred images and low-light conditions, but overall model performance remained stable and satisfactory.
Model Testing and Performance
The trained model was tested using unseen sample images from the testing dataset.
The system successfully:
- Detected price tags in real time
- Generated accurate bounding boxes
- Produced stable detection outputs
- Delivered fast processing speed suitable for real-time applications
Performance Observations
- Fast detection speed
- Good prediction accuracy
- Proper price tag localization
- Stable model performance
- Efficient real-time output generation
The project also included:
- Confusion Matrix
- Precision-Confidence Curve
- Recall-Confidence Curve
- F1-Confidence Curve
- Precision-Recall Curve
for evaluating model performance and detection quality.
Conclusion
Our YOLO-based Price Tag Detection System successfully detected and localized price tags in real time with satisfactory prediction accuracy and fast processing speed.
This project helped us understand:
- Dataset preparation
- Image annotation
- Model training and validation
- Real-time object detection
- Performance evaluation in Computer Vision systems
We also gained practical experience in implementing and testing AI-powered retail detection systems using YOLO object detection techniques.
Demo Video:
(https://drive.google.com/file/d/1Zl0vs9bFi1NF8sS7NrHM4y1Z2LzsRtKk/view?usp=drive_link)
Special Thanks
I would like to express my sincere gratitude to our mentors and faculty members for their constant guidance, encouragement, and valuable support throughout the internship and project development process. Their mentorship helped us better understand the practical applications of Artificial Intelligence and Computer Vision.