Files for the paper: "Sound Source Localization using Deep Residual Learning"
the programs run most on Python (Windows - Linux)
Chainer - Install from pip
Hark - to obtain Audio Features
Run ./chainer_train.py t -C $(config_file) from training folder to train a model
The forwarding file is located in the microcone folder
- Run ./ssl_test.py $(DATE_OF_TRAINED_MODEL) to forward the audio files (Any corpus, any language is fine)
- Run ./compile_results.py to obtain the block accuracy (median angle) - change the exp variable inside the file according to the folder you want to test
- Run ./eval_correc_acc.py to obtain the point-to-point accuracy - change the exp variable inside the file according to the folder you want to test
- dataset_preparation : Two examples of the dataset prepared for the training
- microcone : Files to evaluate any model and a network example to be trained
- python_utils : extra files for training, preparing data, etc.
- training : files for training a network
- training_files : an example of a generated network and the files to test
To generate the impulse use ISM of Eric A. Lehmann
Information of Microcone position microphones at HARK Supported Hardwares
JRM Vol.29 No.1 (Feb. 20, 2017)
This project is licensed under the MIT License - see the LICENSE.md file for details