Watch a live demo of this project https://drive.google.com/file/d/1bax30KmnOahCLjbw7z3pqwct8suVJelR/view
This repository aims at detecting live cough/sneeze with mic . It also detects all other sounds included in ESC-50 dataset
Thanks https://github.com/hasithsura/Environmental-Sound-Classification and https://github.com/karolpiczak/ESC-50
Get the .pth file here https://drive.google.com/file/d/1vba2Fttk4JNssZvUR-p0JhFLsCIURsWA/view?usp=sharing
If you want to train dataset on your own try https://github.com/hasithsura/Environmental-Sound-Classification and for dataset https://github.com/karolpiczak/ESC-50
This repository aims just in giving the inference model not required for you to train!! You of course need a Nvidia CUDA enabled GPU for inference.
To use this program ,
- Run rec_2_db.py and try coughing or sneezing.
- After some seconds rec_2_detect.py starts running and detects whether its a cough/sneeze or something else. (NOTE: The project is not 100% accurate and is fairly slow and need to be optimized for quicker detection and inference)
Libraries required: 1.Pytorch (1.4.0+cu92) 2.Librosa 3.pyaudio 4.numpy and also you require a microphone!!!
NOTE: Specify the paths correctly as mentioned in the program for proper accessing of required file by the program