Get Started

1. Create a virtual environment

python3 -m venv venv

2. Activate the virtual environment

source venv/bin/activate

3. Install the requirements

pip install -r requirements.txt

Generate Spectrograms

The notebook Generate_Spectrograms.ipynb is used to insert gunshots into background sounds.

First the background sounds are loaded from /repo_dir/input/background. Then each background is split into 10 second segments (or windows).

For each window, a gunshot from /repo_dir/input/gushot is inserted at a random position. A given window has one gunshot from three different volume variations: 8%, 35% and 70% volume. There is a 20% chance that a window will have a non-gunshot inserted instead of a gunshot.

Gunshot (or non-gunshot) sounds are always picked at random.

The output spectrograms are saved in /repo_dir/output/spectrograms.

Classification labels are stored in /repo_dir/output/labels.csv, which is a CSV file with the following columns: file_name,label. We can load this file using a custom torch.utils.data.Dataset class, and use it to create a dataloader for training.

Output filenames are of the form:

{backgroundName}_window={index}_vol={volumeLevel}_gun={0|1}.png

Examples:

park_window=7_vol=10%_gun=1.png
crowd_window=1_vol=50%_gun=0.png

Train Models

The notebook Train_Models.ipynb is used to train models. Double_Head_EffNetV2_Training.ipynb has the dual headed Effnet model and training scripts.