prevo 0.19.1

Record arbitrary sensors periodically in an asynchronous manner. Control their properties in real time from CLI. Graph/view tools to visualize live data/images are also provided.

About

Periodic REcording and Visualization of (sensor) Objects

This package provides classes to rapidly create interactive data recording for various applications (e.g. recording of temperature, time-lapses with cameras etc.).

Sensors are read in an asynchronous fashion and can have different time intervals for data reading (or be continuous, i.e. as fast as possible). Synchronous recording is also possible (although not the main goal of this package) by defining a super-sensor object that reads all sensors (and which is itself probed at regular intervals).

Tools for graphical visualizations of data during recording are also provided (updated numerical graphs, oscilloscope-like graphs, image viewers for cameras etc.)

The package contains various modules:

• prevo.record: record sensors periodically, CLI interface, trigger GUI tools from CLI,

• prevo.control: control device properties, create pre-defined temporal evolutions of settings for sensors, devices and recording properties,

• prevo.plot: plot numerical data in real time (regular plots, oscilloscope-like graphs etc.),

• prevo.viewers: live view of images from camera-like sensors,

• prevo.csv: read / save data with CSV/TSV files

• prevo.parser: parse command line arguments to trigger functions or class methods

• prevo.measurements: additional tools to format measurements for Record-like classes.

• prevo.misc: miscellaneous tools, including dummy sensors and devices.

See Jupyter notebooks in examples/ and docstrings for more help.

Below are also minimal examples showing implementation of periodic recording and image viewers.

Install

pip install prevo

Record sensors periodically

For using the package for asynchronous recording of data, three base classes must/can be subclassed:

• SensorBase
• RecordingBase (children: NumericalRecording, ImageRecording)
• Record (children: NumericalRecord, ImageRecord)

A minimal example is provided below, to record pressure and temperature asynchronously into a CSV file, assuming the user already has classes (Temp, Gauge) to take single-point measurements (it could be functions as well). See examples/Record.ipynb for more detailed examples, including periodic recording of images from several cameras.

1. Define the sensors

   from prevo.record import SensorBase
   
   
   class TemperatureSensor(SensorBase):
   
       name = 'T'
   
       def _get_data(self):
           return Temp.read()
   
   
   class PressureSensor(SensorBase):
   
       name = 'P'
   
       def _get_data(self):
           return Gauge.read()
   

   Note: context managers also possible (i.e. define __enter__ and __exit__ in Sensor class) e.g. if sensors have to be opened once at the beginning and closed in the end.

2. Define the individual recordings

   from prevo.record.numerical import NumericalRecording
   
   # Because timestamp and time uncertaintyare added automatically in data
   # Can be renamed to have different time column titles in csv file.
   time_columns = ('time (unix)', 'dt (s)')
   
   # Note: NumericalRecording can also be subclassed for simpler use
   # (see examples/Record.ipynb Jupyter notebook)
   
   recording_P = NumericalRecording(
       Sensor=PressureSensor,
       filename='Pressure.csv',
       column_names=time_columns + ('P (mbar)',),
   )
   
   recording_T = NumericalRecording(
       Sensor=TemperatureSensor,
       filename='Temperature.csv',
       column_names=time_columns + ('T (°C)',),
   )
   

   NOTE: the recordings are already usable as is if one does not need fancy functionality. For example, one can start a periodic recording of pressure every two seconds with

   recording.start(dt=2)
   

   And the following methods and attributes are available:

   # can be set to True or False to pause/resume recording
   recording.active
   
   # can be set to True or False to pause/resume saving to file
   # (e.g. to continue live view of data without saving)
   recording.saving
   
   # Equivalent to playing with .active:
   recording.pause()
   recording.resume()
   
   recording_P.stop()
   # After a stop(), it is possible to call start() again.
   
   # To change the time interval between data:
   recording.timer.interval = ...
   
   # To record data as fast as possible:
   recording.continuous = True
   

   NOTE: recording.start() is non-blocking, so several independent recordings can be started at the same time, and can be changed in real time with the methods/attribute above. However, a convenient interface to manage several recordings simultaneously is provided by the Record class, see below.

3. Record interface for managing simultaneous recordings

   Record provides a real-time CLI to manage recordings properties in real time as well as live data visualization.

   from prevo.record.numerical import NumericalRecord
   
   recordings = recording_P, recording_T
   record = NumericalRecord(recordings)
   record.start(dt=2)  # time interval of 2 seconds for both sensors
   

Many other options and customizations exist See docstrings for more help and examples/Record.ipynb for examples.

Live image viewer

Let's assume one gets image data in a queue from a camera. Individual elements from the camera.queue are dictionaries with keys image (numpy-like image array) and num (image number, an integer). This formatting of elements in the queue is assumed by default, but others are possible (see examples/Viewers.ipynb).

It is possible to view the images using either tkinter, opencv or matplotlib. Here we will use tkinter. One first has to define a window in which to display the images, and then a viewer to operate and update the window.

from prevo.viewers import TkWindow, TkViewer

# The window will show the (display) fps and the image number in real time
window = TkWindow(camera.queue, show_fps=True, show_num=True)

# It is possible to run several windows in parallel from other image sources
# Here we have just one.
viewer = TkViewer(windows=(window,))
viewer.start()

Here again, various options and customizations are possible, see some example in examples/Viewers.ipynb.

Misc. info

Module requirements

Packages outside of standard library

(installed automatically by pip if necessary)

• tqdm
• tzlocal < 3.0
• oclock >= 1.3.2 (timing tools)
• clivo >= 0.4.0 (command line interface)
• gittools >= 0.6.0 (metadata saving tools)
• matplotlib >= 3.1 (due to cache_frame_data option in FuncAnimation)
• numpy

Optional packages

• pandas (optional, for csv loading methods)
• opencv-python (optional, for specific camera viewers)

Python requirements

Python : >= 3.6

Author

Olivier Vincent

([e-mail])