Gunshot-Detection-in-Audio
所属分类:内容生成
开发工具:Jupyter Notebook
文件大小:3842KB
下载次数:0
上传日期:2020-02-16 16:03:56
上 传 者:
sh-1993
说明: 使用TensorFlow 2.0检测枪击的音频分类深度学习模型。测试集准确率97.5%,跟踪率99%...
(Audio classification deep learning model using TensorFlow 2.0 to detect Gunshots. 97.5% test set accuracy and 99% training set accuracy was achieved on Binary-Urban8K. This work was done during my summer internship at TUKL-NUST lab.)
文件列表:
Backups (0, 2020-02-17)
Backups\best_weights_final.hdf5 (203936, 2020-02-17)
Backups\best_weights_modified.hdf5 (203920, 2020-02-17)
Backups\best_weights_temp.hdf5 (580648, 2020-02-17)
Backups\dataframes_backup.h5 (3354024, 2020-02-17)
Misc Scripts (0, 2020-02-17)
Misc Scripts\.ipynb_checkpoints (0, 2020-02-17)
Misc Scripts\.ipynb_checkpoints\0. Expanding UrbanSound8K-checkpoint.ipynb (5351, 2020-02-17)
Misc Scripts\.ipynb_checkpoints\1. Making Customized UrbanSound8K Binary-checkpoint.ipynb (5612, 2020-02-17)
Misc Scripts\.ipynb_checkpoints\2. Expanding US8K-Binary-checkpoint.ipynb (5008, 2020-02-17)
Misc Scripts\.ipynb_checkpoints\3. Expanding Folds in US8K-Binary-checkpoint.ipynb (4917, 2020-02-17)
Misc Scripts\0. Expanding UrbanSound8K.ipynb (5351, 2020-02-17)
Misc Scripts\1. Making Customized UrbanSound8K Binary.ipynb (5612, 2020-02-17)
Misc Scripts\2. Expanding US8K-Binary.ipynb (5008, 2020-02-17)
Misc Scripts\3. Expanding Folds in US8K-Binary.ipynb (4917, 2020-02-17)
# Gunshot-Detection-in-Audio
**Note:** Code for RNN model & audio synthesis is not opensourced yet.
Audio classification using deep learning implemented using TensorFlow 2.0 to detect Gunshots. 97.5% test set accuracy and
99% training set accuracy was achieved on Binary-Urban8K. This work was done during my summer internship at TUKL-NUST lab.
Due to proper preprocessing & feature extraction, a simple CNN model is used to achieve promising results.
## Project Includes
- Project Notebook
- Binary_Urban8K dataset visualization
- Preprocessing and feature extraction using Librosa library
- Pipelines for Preprocessing
- Training using Keras, TensorFlow 2.0
- Predictions on:
- Test files
- Multiple selected files
- Real-time sound input
- MISC Scripts Directory: it includes all the notebooks used to modify Urban8K dataset.
- Backups Directory: it contains *model weights* and *stored dataframe having features extracted from audio.*
## Dataset Details
**Note:** scripts used to modify the data are also provided in the MISC Scripts directory.
- UrbanSound8K was extended by adding 2400 gunshot files to it from AudioSet & MIVIA audio events data set.
- "UrbanSound8K.csv" was modified accordingly.
- 74 more gunshots were added which were downloaded from:
http://soundbible.com/tags-gun.html
- "UrbanSound8K-modified.csv" was created for latest version of dataset.
- "US8K-Binary" refers to new dataset
- Moreover, UrbanSound8K was changed for binary classification with new classes:
- no_gun_shot (8358 files, which is 3 times when compared with other class.)
- gun_shot (2848)
- Total files are 11206.
- Finally, folds in dataset were increased from 10 to 40 to make it work on computers with less RAM memory.
## Other Details
- Currently, dataset is missing from the repository but a download link will be added soon.
- Dataframes having extracted features of dataset files are saved as `dataframes_backup.h5` in `Backups` sub-directory.
- Model with best results is also saved in `Backups` folder named as `best_weights_modified.hdf5`.
近期下载者:
相关文件:
收藏者: