Team members :-
Suraj Gupta, Amitkumar Mishra,Anuj Kurmi
Sahil Khanvilkar
**Use Case Importance"
The dumbbell detection model is important because it enables automated fitness tracking, improves exercise safety through posture analysis, and demonstrates real-world implementation of object detection techniques. It also serves as a scalable foundation for building advanced AI-powered fitness and monitoring systems.
Data collection and Annotation
A dataset of 1500 dumbbell images was collected under various real-world conditions, including different lighting, backgrounds, orientations, and occlusions. The images were annotated using bounding boxes to label dumbbells accurately. The dataset was then divided into training, validation, and testing sets to ensure effective model training and evaluation. This step is crucial for improving the accuracy and robustness of the dumbbell detection model.
Model Training and Validation:
We trained a YOLOv8 model for dumbbell detection using a labeled dataset with various real-world conditions. Key metrics such as precision, recall, mAP, and loss were monitored. Performance was evaluated using confusion matrices, PR curves, and F1-score graphs to ensure accurate detection.
Model Deployment and Demo:
The trained dumbbell detection model was successfully deployed on the YOLOvX mobile app. It provides real-time detection with good accuracy and smooth performance. The best-performing model was selected based on evaluation metrics.
Demo Video Link:
demovideo
Conclusion:
The model performed well in detecting dumbbells in real-time scenarios. Performance improved with a diverse dataset including different lighting, backgrounds, and object positions. Some challenges like blur and occlusion remain, but overall the model shows reliable results and practical usability.