DESCRIPTION OF THE DATASET
This dataset was created for real-time fitness exercise classification and includes a diverse mix of synthetic and real-world videos. It focuses on four common exercises:
- Squat
- Push-up
- Barbell Bicep Curl
- Shoulder Press
DEMO OF MY PROJECT THAT USED THIS DATASET (AI PERSONAL TRAINER):
GITHUB PROJECT (with implementation): https://github.com/RiccardoRiccio/Fitness-AI-Trainer-With-Automatic-Exercise-Recognition-and-Counting The dataset was compiled from three main sources:
📁 Kaggle Workout/Exercises Video Dataset
- Real-world videos of expert trainers performing various exercises
- Only four exercises were selected
- ~25 videos per class were curated, ensuring balanced representation
- Supplemented with additional online videos to increase variation in lighting, angle, and environment
🧍 InfiniteRep Dataset
- Synthetic videos of human-like avatars performing exercises
- 100 videos per class selected
- Offers control over pose variation, camera angles etc.
- Enhances model robustness and dataset size
🌐 Additional Online Videos
- Sourced from Pexels, Pixabay, Shutterstock, etc.
- Added to reflect how users might perform exercises in home or gym environments