python-recorder
Device independent framework for synchronized recording. Originally designed to record audio and video from a Realsense camera and several microphones. This framework is designed for researchers collecting multimodal datasets. It provides an easy to use command line application that allows to configure, test and use the available sensors in the host machine while providing intuitive mechanisms to synchronize the recording devices with the specific event that one wants to record. The framework can be divided in two main features:
Devices
They provide a common interface to initialize, start
and stop
a recording,
whether is a camera, a microphone or any other sensor. When the recording is
stopped, device parameters (e.g., samplerate
in microphones, framerate
in
cameras) as well as a start_timestamp
and an end_timestamp
are saved
together with the recorded data which allows fine synchronization of data from
multiple devices in post-processing.
Devices are configurable through a command line application, which searches for
available options in the host machine. At the same time, they are saved in a
yaml
formatted text file. Which allows easy access
to device exclusive configuration parameters.
Multiple devices compose a Recorder
object, which
is responsible for initializing all its devices, setup an output folder for the
recordings, and start and stop the recording.
Adding new devices classes is simple, one just needs to extend the Device
class and override its abstract methods (find
,
_start
, _stop
and show_results
).
Listeners
Researchers usually want to synchronize the recording with a specific event, a
test or experiment, which can be triggered by a different machine. A Listener
is an interface between a Recorder
and an external event. At the moment, the
only one available is a Localhost
listener. Which triggers the recording
according to GET
requests through the local network. But additional listeners
can be added by extending the Listener
class and overriding its abstract methods. An
example of another listener would be a ROS
node that subscribes to a specific
topic and triggers the recording according to a specific set of messages.
Listeners can be setup through the listen
command. This command does not initialize
the devices. The listener is responsible of executing recorder.setup()
as it
sees fit. In the case of the localhost
listener, a GET
request to the
/setup
endpoint is necessary before trying to start the recording via the
/start
endpoint.
Setup
It is recommended to use a virtual environment in order to isolate the
installation, as it provides a command line interface named recorder
that can
conflict with other system packages.
python -m venv venv
Activate the environment on Windows:
venv\Scripts\activate
or on MacOS and Linux:
source venv/bin/activate
Install from pypi. There are optional dependencies that define which devices and listeners are available, due to the fact some of these dependencies are not available for all operative systems. Check extras for more information about it.
pip install python-recorder[all]
or install this repository to use the latest development versions.
pip install git+https://github.com/AcousticOdometry/python-recorder.git[all]
Extras
Extras can be installed through the pip install
command
by specifying them at the end of the package name:
pip install python-recorder[extra_one,extra_two]
The available extras are defined in pyproject.toml
and
are described here:
-
realsense
: Intel Realsense devices depend on pyrealsense2. -
all
: All the extras defined above.
Usage
Check the usage with the --help
option:
recorder --help
Workflow
Configure the devices to be used. One can always modify the configuration
manually in the generated yaml
file.
recoder config
Test that the chosen audio devices are working
recorder test microphone
Record an experiment with the configured devices
recorder record
Listen to GET
requests through the local network
recorder listen localhost
Contributing
We are open to contributions. Please, feel free to open an issue or pull request with any improvement or bug report. The framework is simple and easily extensible, it should fit easily in many projects.
TODO
- ROS listener
- Complete test suite
-
Camera device with
OpenCV
- Speedup the configuration file creation with compiled code?