SADACO : Swift Audio DAta COmprehension Framework + Stethoscope Audio DAtaset Collections

Welcome to SADACO repository! Where spooky chest sounds are brought to life. [Docs]

!!! Web Server for deploying sadaco service is available at SADACO-WEB !!!

We currently provide basic train-test-explain pipelines for ICBHI, Fraiwan, PASCAL-Heartsound and bunch of other NN APIs.

**** We are still under development process. Let us know if there's any question or suggestion. We'll try to make it happen :) ****

**** READMEs and Documentations are also getting in shape ****

Most detailed information will be provided in the official document.

Installation

Environment Setting

Using Conda (Recommended)

conda env create -n sadaco -f sadaco_env.yml

Using pip

python -m pip install -r requirements.txt

Installing SADACO

RUN

python setup.py develop

or

python -m pip install -e .

We are now on a development phase. So we recommend installing in a development mode so changes on the source codes can be immediately reflected to the package.

Getting Started

Currently, we provide the training and validating code for only the ICBHI dataset. First move on to the pipelines folder and, RUN

python ICBHI.py --conf-file configs/train_basic.yml

This will start running ResNet50 training on ICBHI official split dataset with Banlanced Batch Sampler.

You can also try different settings by running

python ICBHI.py --conf-file configs/train_contrastive.yml

or changing values in the yaml files, creating new configuration that suits your intention.

---

Contributors ✨

Thanks goes to these wonderful people(rolls shown by icons):

dlrudco 🤔💻📖🚧📦💬🔬

lunayht 🤔💻📖📦💬🔬

This project follows the all-contributors specification. Contributions of any kind welcome!

---

Inspiring References

1. Supervised Contrastive Loss