Pose Instructor
The main aim behind this project is to aid all the atheletes to workout and exercices in the most accurate way possible. For an athlete to get maximum out of the body, they have to do everything right. Doing excercise in the wrong will not benefit the body at all, but will cause alot of injuries. Du
2025-06-28 16:34:31 - Adil Khan
Pose Instructor
Project Area of Specialization Artificial IntelligenceProject SummaryThe main aim behind this project is to aid all the atheletes to workout and exercices in the most accurate way possible. For an athlete to get maximum out of the body, they have to do everything right. Doing excercise in the wrong will not benefit the body at all, but will cause alot of injuries. Due to wrong excercise atheletes can get Muscle pull and strain, Sprained ankle, Shoulder injury, Knee injuries, Shin splint, Wrist sprain and Tendinits. These injuries can cause the athelete career to end or halt for a long time. A study published in 2013 states that there is a rate of 3.1 injuries for every 1,000 hours spent doing CrossFit training.
Our main objective is to rectify the problems an athlete can face while training. This solution provides a virtual assistance to the athletes, tell the athletes how to do an excercise and also check the atheletes correct angles and movements. The solution is going to be a complete package. A professional can teach the machine how a particular excersice is done in the most efficient. After the machine is trained on a particular excercise, the athlete or beginner can workout live and the machine will tell whether its correct or wrong. Not only this the machine will be mature enough to tell where actually the athlete is making mistake and how to correct it.
We are going to create 2 different versions of our product. One for professionals atheletes and trainers where the solution will be deployed on hardware and will be available as a device. A GUI will be available to train, test and doing other related operations. The 2 version will be available on mobile side where user can record his/her video according to the predefined excercise. The video will then be send to server for processing and generating issues.
In short it is a solution which is adaptable and saves you from many injuries.
Pose estimation is a computer vision technique that predicts and tracks the location of a person in an image or video.We can also think of pose estimation as the problem of determining the position and orientation of a camera relative to a given person and we done it by identifying, locating, and tracking a number of KeyPoint's on a given person.We clearly envision the power of pose estimation by considering its application in automatically tracking human movement. From virtual sports coaches and AI-powered personal trainers to tracking movements on factory floors to ensure players safety. Our main purpose to train the players in sports to obtain their angle with respect to their game and helps them to perform well without any injury.Our main target domain is sports,gym,athlete
Project Implementation MethodWe implemented the alpha pose technique to meet our desired results through deep learning. Alpha Pose is latest research paper on 2D pose estimation. We are using Mxnet & Gluoncv for implementation of pose estimation.
1) First we implemented the YOLO model for person detection from frames
2) Then YOLO draws human bounding boxes on frames
3) These human bounding boxes are fed into AlphaPose Algorithm
4) AlphaPose predicts human joints position
5) Then AlphaPose will return 18 body key values
6) Lines are drawn on person’s body through these 18 body key values
7) Then these lines are sketched on frames
8) Then angles between different body parts are calculated
9) If the angles are correct during activity then lines on that body part are shown in green
10) Lines on that body part will turn red if any given angle is wrong on that part of body
We solve this problem and use it in three main domain i.e sports,GYM,athlete. Ones who use this projects can get the benifit not to hire trainer in GYM or if he/she wants to do home workout for their angle correction.to provide safety to the players from any kind of injury happeend due to wrong angle and also help the players to achieve the right angle for the specific exercise or phase.
Technical Details of Final DeliverableThe final deliverables for our project is a document explaining everything about the project, a Mobile application and an IOT device. The document will be explaining everything about the project. It will consist of the purpose and vision why we made this app. Morever, the development process will also be visuziled there with help of different software engineering diagrams.
The mobile application will be available for all Andriod users. We are planing to commercially launch this app aswell. The app will be free for all users. The application will have a clean and guided GUI for all the users. The processing time of a video will be around 30-60 secs depending on the network connection. The application will be designed to handle 1000 request at a time having a crash probability of 1 in 10000.
The IOT device will be available through ecommerce market such as daraz for all the professionals to buy it. The device will contain a user manual which will guide the user throughout the GUI of the device. It will also tell how to setup and efficiently use the device. The device will be powered by a rechargeable battery.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 80000 | |||
| Camera | Equipment | 1 | 6000 | 6000 |
| LED | Equipment | 1 | 15000 | 15000 |
| NVIDIA Jetson Nano | Equipment | 1 | 29000 | 29000 |
| Server | Equipment | 1 | 20000 | 20000 |
| Professional Trainer Fee | Miscellaneous | 1 | 7000 | 7000 |
| Travel + Stationary + Printing | Miscellaneous | 1 | 3000 | 3000 |