Tools

Instead of specifying a visual handler, you may specify a tool as argument for beatviewer:

python -m beatviewer <Tool Name> <Tool Args>

Possible names are annotator, directogram and dummy.

Content

• Directogram Computation
• Video Annotator
• Dummy Tracker

Directogram Computation

The directogram takes a video (or GIF) file as input, computes its directogram and extracts checkpoints that can be used for the warp visualizer. It outputs a JSON file next the video file. It implements a technique described in Visual Rhythm and Beat, Davis, 2018.

Argument	Default	Description
path		Path to the video file
skip-directogram	False	Set this flag to skip directogram computation, and reuse saved data
average-period	1.0	Length in seconds of the window for computating signal average and standard deviation
maximum-period	0.15	Length in seconds of the window for computatig signal local maxima
sigma	2.0	Threshold to detect a checkpoint, as a number of std above the mean

The first step, directogram computation, takes a lot of time, while the second step, checkpoints extraction, requires a lot of finetuning. A practical way to use this then is to run the script once: the output JSON file will contain the directogram. Then run the script again with the -s flag to skip the directogram computation, and tune the checkpoint extraction parameters as much as you want.

Video Annotator

I find the Directogram Computation inneficient: it takes a lot of time to compute and the result is not very accurate. The video annotator script uses Pygame to allow doing this work manually in an very time efficient way. It takes the path to the video to annotate as input, and uses very basic key controls:

Key	Function
Right	Go to next frame
Left	Go to previous frame
Ctrl + Right	Go to next checkpoint
Ctrl + Left	Go to previous checkpoint
Ctrl + A	Gox to first frame
Ctrl + E	Go to last frame
Shift (hold)	Move one frame/checkpoint at a time
Alt (hold)	Move as fast as possible
Space (hold)	Warped playback at 100 BPM
Enter	Toggle checkpoint
Ctrl + S	Save to JSON
Escape	Exit
Ctrl + W	Exit

Argument	Default	Description
path		Path to the video file
size	900	Maximum size of the window
buffer-before	30	Number of frames stored in the buffer before its cursor
buffer-after	30	Number of frames stored in the buffer after its cursor
buffer-margin	120	Number of frames above the default buffer size to trigger the cleanup

Dummy Tracker

To make developping visualizers easier, I added a dummy beat tracker. Give it a target BPM and then pass visualizers arguments as you would for the real beat tracker. The dummy tracker will then regurlarly send fake BPM and BEAT messages down the multiprocessing pipe.

Argument	Default	Description
bpm	—	Target BPM (might not be very precise)
onsets	False	Send onsets every beat and half beat
print	False	Print beat to stdout
beep	False	Make noise when a beat occurs
keyboard	False	Enable keyboard shortcuts (PageUp and PageDown for increasing or decreasing the BPM)