Traditional ML classifiers are easy to understand and implement, fast and provide good explanations of the data and predictions. However, their performance is compromised when the data is high dimensional and large. We have analysed and compared ML algorithms, namely K Nearest Neighbours, Naïve Bayes, Random Forest and Support Vector Machine to classify different yoga pose images into appropriate asanas. We tried to improve the results by techniques like feature extraction and edge detection by using Gabor filters and Sobel kernels. The project describes the limitations of traditional ML classifiers and suggests the need for deep learning techniques to handle this problem more accurately.
Data set : https://www.kaggle.com/niharika41298/yoga-poses-dataset