car-sound-dataset 1.5.0

You can use the car sound dataset with this Python package.

car_sound_dataset

This repo contains the source code of car sound dataset Python package.

https://pypi.org/project/car-sound-dataset/

How to install?

pip install car-sound-dataset

About the dataset

The dataset contains 17121 car sound events. Each event is recorded by at least 1 and maximum 6 devices. The length of each recording is 3 sec. The sample frequency was 3906 Hz during the measurements. You can download the dataset and select different recordings with this package.

Class

    A class used to represent the car sound dataset.

    ...

    Attributes
    ----------
    path : str
        A path, where you want to download and use the dataset.
        The default path is the current working directory.

Methods

    download_and_get_dataset()
        Downloads the dataset and the json file, which identifies the events.

    list_events_of_dict( car_dataset_dict)
        This method prints to console the event IDs of a car_dataset_dict.
        
    plot_event(car_dataset_dict, event_id)
        You can plot an event with this method.

    play_event(car_dataset_dict, event_id, nomralize=True)
        You can play the recordings of an event with this method.

    filter_by_pattern(car_dataset_dict, pattern)
        This method helps you to select events by a pattern.
        The pattern should contain 'X', '0' or '1'.
        'X' - You don't care, if the device recorded the event or not.
        '0' - You skip the recordings of this device.
        '1' - This device recorded the event.
        You have to give the pattern as a list. Each list element represents each device.
        An example pattern: ['0', '1', 'X', '0', 'X', '1']
        This means we skip the recordings of DEVICE1 and DEVICE4. We select the events,
        where DEVICE2 and DEVICE6 recorded the event. Each event will contain 2-4 recordings.
        It depends on DEVICE3 and DEVICE5. If one of DEVICE3 and DEVICE5 recorded the event,
        the event will contain 3 recordings and if both is recorded the event, it will contain
        4 recordings.

    filter_by_exact_device_count(car_dataset_dict, device_count)
        This method selects the events, which is recorded by the given count of devices.

    filter_by_min_device_count(car_dataset_dict, device_count)
        This method selects the events, which is recorded by at least the given count of devices.

    load_dict_as_pool(car_dataset_dict)
        This method loads the recordings into a 3D numpy array. In this case each element
        contains only one recording. The recordings of an event are split to single recordings.

    load_dict(car_dataset_dict, is_random = False, device_indexes = [])
        This method loads the recordings into a 3D numpy array. In this case each element
        contains events. One event usually contains more than one recordings.