In this project, I am classifying 10 different sounds. This project was inspired by Krish Naik's YouTube video. The dataset is publically available from here
Steps
- Download the dataset
!wget https://zenodo.org/record/1203745/files/UrbanSound8K.tar.gz
- Go to the directory of downloaded dataset and extract using
!tar -xf <file_name>
- To train the model, use command as,
!python train.py --checkpoint <location where you want to save the weights> --epochs <epoch number> --dataset_dir <directory of dataset>
Implementation of predict.py is not complete yet.
The accuracy obtained by Sound_Classifier.h5 is 92% on the test data.