In this period of the pandemic, when it is not safe to stay outside our houses,
Our application aims at helping users to improve their skills as well as help them maintain their fitness levels.
People can use this tech and enjoy learning at home without any hassle while maintaining social distancing,
It's a fun way as we can see our body movements with precision and help us improve if there is some scope of improvement.
This is a lightweight app designed for ease of usage and accessibility
Cosine Similarity
In order to grade the user, we have to check the cosine similarity between the body movement of the user and the instructor.
Cosine similarity demands two vectors as input, each one having the x and y coordinates of 17 body parts for each frame that is encountered. We weren't able to find an effective way to create these vectors.
Problem 1- the user video and the instructor video needed to be of equal length.
Problem 2- the user has to start dancing as soon as the recording starts.
Problem 3- creating a general solution for this was tough, as more than one pose was detected for a person (one having higher confidence and the other having lower confidence).
Technologies used
Discussion