India is a land of diversities and we have an array of different dance forms that are popular in different parts of our country. This project aims to develop a deep learning model to identify a dance form from an image. The model can identify 8 major dance forms (Mohiniyattam, Odissi, Bharatanatyam, Kathakali, Kuchipudi, Sattriya Nritya, Kathak and Manipuri Raas Leela) with an accuracy of 67% as of now.
The model is a fine tuned version of the pretrained convolutional neural network, VGG16. The fine tuning was done with a relatively small dataset of 364 images from this dataset. The model is deployed with Streamlit. You can access it here.
For more detailed discussion on the model building process. Please refer to this notebook.