Skip to content

Instantly share code, notes, and snippets.

@Zulko
Last active December 15, 2020 17:36
Show Gist options
  • Star 6 You must be signed in to star a gist
  • Fork 2 You must be signed in to fork a gist
  • Save Zulko/633c1b0807b37aa52d9c to your computer and use it in GitHub Desktop.
Save Zulko/633c1b0807b37aa52d9c to your computer and use it in GitHub Desktop.
Growing bacteria with population size analysis (Matplotlib/MoviePy)
"""
This script produces a video analysis proving for the first time in history
that bacterial populations grow exponentially. See the result here:
http://i.imgur.com/uoITKiA.gif
We proceed as follows:
- Download a video of bacteria growing under a microscope.
- Cut the video to keep only the first 7 seconds.
- Threshold each frame to find where the bacteria are, and compute the
total number of pixels that they occupy
- Make an animated plot of the number of pixels occupied versus time
- Assemble the original video, the thresholded version, and the plot,
and write everything to a file.
I don't know the source of the original video so I can't give credits,
sorry guys !
"""
import numpy as np
import matplotlib.pyplot as plt
from moviepy.video.io.bindings import mplfig_to_npimage
from moviepy.editor import VideoFileClip, VideoClip, clips_array
# DOWLOAD THE VIDEO Requires Youtube-dl installed.
import os
os.system("youtube-dl gEwzDydciWc -o growth.mp4") # requires Youtube-dl
# LOAD THE VIDEO, SELECT THE EXCERPT BETWEEN t=0-7 seconds
video = VideoFileClip("growth.mp4", audio=False).subclip(0,7)
# THRESHOLD THE FRAMES TO GET THE POPULATION AREA
# We normalize each frame by the luminosity of its upper left corner
# to correct for the changing luminosity of the video
thresholder = lambda im: ( 1.0*im[:,:,1]/np.mean(im[:50,:50]) ) < 0.8
thresholded = video.fl_image(thresholder)
thresholded_RGB = thresholded.set_ismask(True).to_RGB() # RGB Version for movie
areas = [(t,frame.sum())
for (t,frame) in thresholded.iter_frames(with_times=True)]
# ESTIMATE THE GROWTH RATE
tt, aa = [np.array(e) for e in zip(*areas)] # times and areas
gr_rate, a0 = np.polyfit(tt, np.log(aa),1)
# PLOT THE FIGURE
fig, ax = plt.subplots(1, figsize=(3,2.5), facecolor=(1,1,1))
ax.plot(tt, np.exp(a0+gr_rate*tt), lw=2, ls='--', c='k', label = r"$Ae^{rt}$")
l, = ax.plot(tt, aa, lw=3, c='r')
ax.set_ylabel("Pop. surface (pixels)")
ax.set_xlabel("time (unit unknown)")
ax.legend(loc=2)
fig.tight_layout() # <- I love this function :D
# ANIMATE THE FIGURE WITH MOVIEPY
def plot_until_t(t):
tt, aa = zip(*[(t_,v) for (t_,v) in areas if t_<=t])
l.set_xdata(tt)
l.set_ydata(aa)
return mplfig_to_npimage(fig)
plotclip = VideoClip(plot_until_t, duration=video.duration)
# ASSEMBLE ALL THE CLIPS, WRITE TO A FILE
final_clip = clips_array([[clip.margin(2, color=[255,255,255]) for clip in
[video.resize(.6), thresholded_RGB.resize(.6), plotclip]]],
bg_color=[255,255,255])
final_clip.write_videofile('growth2.mp4', fps=15)
#final_clip.write_gif('growth.gif', fps=15, opt="OptimizeTransparency", fuzz=10)
@Zulko
Copy link
Author

Zulko commented Feb 17, 2015

@BaddMann
Copy link

Any chance this would work with GreenScreen?
I'm failing to find any moviepy green screen tips....

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment