tony32769 / bokeh_light_curve_inspector

A simple pure-python web browser application using Bokeh to visualize, inspect and label periodic astronomical time series

Bokeh Light Curve Inspector

A simple pure-python web browser application using Bokeh (bokeh.pydata.org) to visualize, inspect and label periodic astronomical time series. Possible use-cases are novelty detection, collaborative research and crowd-sourcing. Disclaimer: This code was written for educative purposes and it is not safe for production (see TODO below).

How to run

cd this_folder
mkdir example/results
bokeh serve . --args example/light_curves example/periods.pkl example/results

Then point your favorite web browser to

http://localhost:5006

Note about the arguments

The Bokeh server allows to pass arguments to the main code using the --args flag. Three arguments are expected:

1. Path to the folder containing the light curve files
2. Path to a pickle file containing file names and light curve periods
3. Path to a folder to dump results from labeling process

Reading and parsing of the light curve files is done in light_curve_handler.py

What you should see when running the example

Log-in interface:

Inspector interface:

How does it work

• Choose a username and hit enter at the log-in interface
• You should be seeing the inspector interface. Here you can:
  • Move from page to page, each showing a set of 9 light curves
  • Label the currently selected light curve into one of the available categories
  • Each time you move to the next/previous page your labeling results are saved at resultfolder/username/

Remote access to the bokeh server

If you intend to access the bokeh server from another machine in your LAN you will have to add

--allow-websocket-origin=myserverlanip:5006

before --args in the execution line above

If you intend to access from outside your LAN just change myserverlanip by the desired domain. Bokeh serve also accepts a --port option if you need to use a different port

Requirements

1. Numpy (numpy.org)
2. Bokeh (bokeh.pydata.org)

Note: Tested with Python 3.6 and Bokeh 0.12.6

TODO

• Add test data
• Improve server security to make this safe for production
• Add user authentication, sessions, etc

License

MIT