Smart Gym Trainer Using AI Pose Estimation
As the pandemic has hit the world and close human contact has become dangerous, people are getting isolated at their homes and almost every activity is now done in private space. In this situation fitness center are effected the most. In order to tackle this problem, we will be buildin
2025-06-28 16:29:19 - Adil Khan
Smart Gym Trainer Using AI Pose Estimation
Project Area of Specialization Artificial IntelligenceProject SummaryAs the pandemic has hit the world and close human contact has become dangerous, people are getting isolated at their homes and almost every activity is now done in private space. In this situation fitness center are effected the most. In order to tackle this problem, we will be building a personal gym trainer based on computer vision and deep learning technologies. However, we also observed that People doing exercises on their own are more susceptible to wrong posture and angles, In our project, we are using Pose estimation techniques that detects the user’s exercise pose and angles and provides real time feedback of his Pose, angles and reps. Through which user can improve their form. Gym Trainer utilizes the best in class pose estimation to distinguish a user's pose by using pose landmarks, then evaluates the vector geometry of the posture through an activity to give valuable feedback.
Project ObjectivesTo reach the outcome of a smart gym trainer application the following are our goals:
1. The software system will use machine learning to learn and estimate a variety of human poses . For this purpose we will require a large data set in order for the software to be as accurate as possible in pinpointing the irregularities in a particular exercise.
2. For the system to set a baseline with which it will compare all data inputs like various angles of exercise we will need the data samples of professional bodybuilders/athletes. Their workouts will be considered ideal and be compared with the average persons workout.
3. The graphical user interface will need to be as user friendly as possible so that the user would have minimal problems in uploading his data like his images or video during a workout.
4. After comparison between the average and ideal workouts the software will have to output data in the form of a graph displayed on the screen of a persons digital device. In addition to this the application will have to provide an adequate feedback to the user so he can adjust his workout accordingly.
Project Implementation Method
1. The software system will use machine learning to learn and estimate a variety of human poses. For this purpose, we will require a large data set in order for the software to be as accurate as possible in pinpointing the irregularities in a particular exercise.
2. For the system to set a baseline with which it will compare all data inputs like various angles of exercise we will need the data samples of professional bodybuilders/athletes. Their workouts will be considered ideal and be compared with the average person’s workout.
3. The graphical user interface will need to be as user-friendly as possible so that the user would have minimal problems in uploading his data like his images or video during a workout.
4. After comparison between the average and ideal workouts the software will have to output data in the form of a graph displayed on the screen of a person’s digital device. In addition to this, the application will have to provide adequate feedback to the user so he can adjust his workout accordingly.
Technical Details of Final DeliverablePersonal Gym trainers’ apps are long time desire of the people who don’t want to go gym and perform workout in their ease, this is not a new problem but recently it gets highlighted due to pandemic. Not much work is done in this area so very few solutions are present in the market.
Convolutional Neural Networks are widely used in this problem to train image data set to detect joints of a human body.
OpenPose and DeepCut are another Deep learning model which is used in multi person pose estimation.
Disadvantage of using other methods is that nearly every other method are relatively slow, and they are very computationally expensive and require a lot of hardware and software capability.
Final Deliverable of the Project Software SystemCore Industry ITOther Industries Medical Core Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Good Health and Well-Being for PeopleRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 9050 | |||
| Webcam | Equipment | 1 | 5200 | 5200 |
| hdmi cable | Equipment | 1 | 350 | 350 |
| poster | Miscellaneous | 1 | 2500 | 2500 |
| Printing | Miscellaneous | 200 | 5 | 1000 |