Running TensorFlow with Python 3 kernel in Docker on RaspberryPI

Installing docker on Raspberry PI 3

Running TensorFlow

With Docker installed on the Raspberry PI running TensorFlow is this easy:

docker run -d -p 8888:8888 mfandreas/rpi-tensorflow-python3

Now hit your browser to


Using shotwell to label images

I needed to label a lot of images into 4 classes. Using the ratings of shotwell this can be really fast.

Pressing 1 to 5 to set a rating on an image.


The next step was to get the ratings for later use as labels in machine learning.

Because shotwell stores its data in a sqlite database we only need some lines of Python to get the labels:

import csv
import os.path
import sqlite3
import sys

# location of shotwell sqlite file
db_file = os.path.join(os.environ.get("HOME"), '.local/share/shotwell', 'data', 'photo.db')
conn = sqlite3.connect(db_file)
c = conn.cursor()
rows = c.execute('SELECT filename, rating FROM phototable WHERE rating > 0 ORDER BY filename;')
csvwriter = csv.writer(sys.stdout, delimiter=';')
csvwriter.writerow(["filename", "rating"])
for (filename, rating) in rows:
    csvwriter.writerow([os.path.basename(filename), rating])

Only the ratings bigger than 0 are selected. The CSV is written to stdout by default.

Kindle 4 as information display

I used a Kindle4 as status display some time ago. For some months this display was without power and I was to lazy to fix it. My solution in 2014 rooting and image generation is nowadays too complicated.

Ciko from my local hackerspace solved the problem in an easy way using the buildin webbrowser of the Kindle 4. No rooting required.

Repository of Cikos solution:

My kindle today:


Disable screensaver (source):

;debugOn (press enter)
~disableScreensaver (press enter)