https://github.com/ashish10alex/web-drone-project
This repository is inspired from Webmushra and Pymushra modified for AB testing
https://github.com/ashish10alex/web-drone-project
flask webmushra
Last synced: about 2 months ago
JSON representation
This repository is inspired from Webmushra and Pymushra modified for AB testing
- Host: GitHub
- URL: https://github.com/ashish10alex/web-drone-project
- Owner: ashish10alex
- Created: 2021-05-15T17:12:53.000Z (about 5 years ago)
- Default Branch: main
- Last Pushed: 2021-06-29T09:31:31.000Z (about 5 years ago)
- Last Synced: 2025-08-30T05:02:13.943Z (10 months ago)
- Topics: flask, webmushra
- Language: Jupyter Notebook
- Homepage:
- Size: 300 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# web-drone-project
This repository is inspired from [Webmushra](https://github.com/audiolabs/webMUSHRA) with [Pymushra](https://github.com/nils-werner/pymushra) which provides a python based backend to store results.
This project has following key modifications from above repository to accomodate it for our AB testing experiments.
* Completely port to Flask based development.
* No need for separate WebMushra directory - Javascript dependencies have been added to `static` folder
* Key entry point for web app assuming `web-drone-project/pymushra/pymushra` as root is is -> `service.py -> templates/index.html -> static/startup.js`
Key Experiment details -
* SNRs - `['0', '-5', '-10', '-15', '-20', '-25', '-30']`
* Number of unique noisy utturances for each model and each SNR - 15
* Number of pairs / pages each participant will see - 20 [controlled by `window` param in `generate_yaml_full.ipynb` and `generate_yaml_full-noisy_baseline_combinations.ipynb`]
* Total number of pairs -
* Baseline vs Input - 105
* Remaining algorithms - 630
Audio files for this experiment are in -
```
configs/resources/Drone_Noise_Test_Data/drone_noise_out/
.
├── Clean
├── DCUNet
├── DPTNet
├── Noisy
├── RegressionFCNN
├── SMoLnet
└── WaveUNet
```
## Setup steps
Please skip to bullet point 2 as I have aleady generated these meta data files.
1. Meta data Genration
If you want to generate your own meta data you will need to change paths accordingly. The current paths are based on paths from Jade and my personal PC
`meta_data.ipynb` - Generate CSV paths for audio files for various SNRs and mdoels
`generate_yaml_full.ipynb` - Yaml files for remaining 4 algoritms `['DCUNet', 'DPTNet', 'SMoLnet', 'WaveUNet']`
`generate_yaml_full-noisy_baseline_combinations.ipynb` - Yaml files for Baseline model which will be compared to Input only.
2. Run the server locally
```
git clone https://github.com/ashish10alex/web-drone-project.git
cd web-drone-project
python3 -m venv .
source bin/activate
pip install -e pymushra
pymushra server
```
Then open `http://localhost:5000`
3. Database
* But the main database which we will use for evaulating results is in the root of the repository - `database_baseline_vs_noisy.csv` and `database_other_model_combinations.csv`
* To view the results for each subset can be viewed by switching btw dropdown menus - [Error handling TBD]
* `http://localhost:5000/result`
* To see all the results for both sets of experiments go to - `http://localhost:5000/admin` and click on `preview`
4. Monitor Experiments
* You can monitor the finished and remaining experiments at - `http://localhost:5000/finished`