-
Notifications
You must be signed in to change notification settings - Fork 250
Home
The GUI is intended as a launcher for Kilosort, to help you run it on your data without errors and to do some sanity checks that it is working properly. It does NOT include any functionality for manually curating/revising/editing the results of the algorithm, so it is NOT a replacement for Phy or for other careful manual inspection.
To install, first follow the instructions at the main page.
Next, navigate to the kilosort directory and run 'kilosort':
>> cd \my\kilosort2\directory\
>> kilosort
To get started, choose a dataset by clicking "select data file" in the upper left. The data file should be a raw binary file containing just the data matrix, size [nChannels x nTimepoints], with int16 datatype. For more on data files, see below.
After you select the file, Kilosort will try to load it using the default channel map and
The Probe View (middle) shows the layout of your probe as specified in the channel map. Green/yellow channels are ones selected currently for viewing in the Data View. Blue/red channels are not currently visible in Data View. Yellow/Red channels are excluded from analysis, and do not appear in Data View.
The Data View shows the data from your file as kilosort will see it. If you can't see spikes, there's a problem! See troubleshooting.
In colormap-mode, the Data View has an image showing channels (y-axis) by time (x-axis) voltage values. In this mode, only one of the available datasets (raw, filtered, prediction, residual) can be seen at a time.
In traces-mode, the Data View shows a set of traces for the selected channels, spaced out along the y-axis. In this mode, the different available datasets can be overlaid.
The datasets are:
- Raw data: the data straight from your file
- Filtered: after preprocessing (whitening, common average referencing, temporal filtering) as specified in the options.
- Prediction: based on Kilosort's results, what does it think the 'true' data looked like. Ideally, it should be a de-noised version of the filtered data.
- Residual: the difference between Filtered and Prediction, i.e. anything Kilosort didn't get right.
Here is a detailed list of controls available:
-
Click
main data plot - re-center the view on that timepoint/channel -
Click
on the grey timeline bar (bottom) - jump to that timepoint -
c
- Flip between colormap-mode and traces-mode -
Scroll up/down
- advance forward/backward in time -
Ctrl
+Scroll up/down
- make the data view contain fewer/more channels of data -
Shift
+Scroll up/down
- zoom in/out in time -
Alt
+Scroll up/down
- change the data scaling. In colormap mode, this means the color scale. In traces mode, it means the height/y-axis scaling. -
1
/2
/3
/4
- In colormap mode, flip back and forth between raw/filtered/predicted/residual datasets. In traces mode, each number toggles one of these datasets.
-
Click
- select channels to view that are nearest to where you click -
Scroll up/down
- zoom in/out on the probe view -
Right-click
on a channel - toggle whether this channel is enabled/disabled for analysis
Click "Run all" in the bottom left. Note that while it is running, the gui will mostly be un-responsive. You will get status messages in the log box, and more detailed messages at the command window. Figures may appear.
Try clicking somewhere (returning the focus to the window) and try again. Or, be patient for it to finish computing something. If not that, close and re-open the gui.
Try resetting the GUI with the Reset GUI button at upper right and starting over. No? Create an issue with a detailed description.