iPhone 8 barbell squat form measurments

About the dataset This data was obtained using Iphone8. Created a Swift program that can access sensors like: accelerometer, gyroscope and other values. Magnometer was not included because a bunch of sources told that it was dependent on the surroudings. If you have a Iphone and a Mac you can use Xcode to measure squats. IMUREADER - https://github.com/likaiZnazis/IMUReader. Keep in mind that there is a sensor that is best suited for segmenting the signal (finding where the individual squats were made). Datasets.zip file contains two datasets: 50hz and 100hz sampling rate. Meaning the sensor registered changes every 20 or 10 ms. Both 50 and 100 hz folders contain a bunch of csv files. These csv files where raw squat data is. Their names correspond to what type of form was performed, each csv file contains 10 squat segments. Total amount of squats per dataset are 240. Every single class is balanced both trough timestamps and segment amount. 100hz sampling rate is much better than 50hz. How the dataset was made You can view the full code of how this dataset was made at https://github.com/likaiZnazis/squat-Posture. First you will probably want to install all the dependency's, nice that I stored them inside a file - requierments.txt. Matplotlib must be installed also. It's saved inside a numpy file. It's dimensions are (segments, features, timestamps). You can view this by: import numpy as np import os data1 = os.path.join(os.getcwd(),"dataset","hz100","train_dataset.npy") data = np.load(data1) print(data.shape) Making your own dataset You can use code from both repos to make your own forms and squats. Keep in mind the phone placement (I placed it on the right thigh). By running command-line.py it will give you a interface on how you should run your commands to make a quality dataset.