Gathering Training Images with the Raspberry Pi

The following scripts were used to gather overhead images of an office for the purpose of training a CNN to detect occupancy within that office.

---

The timed_capture.py script captures images at a fixed rate and saves them within a directory.
Usage:

• Supply the script with a folder path as a cmd-line argument to change where images are saved.
• Modify the captureRate variable to change the period at which images are taken, in seconds.
• Modify the endTime to change when the script automatically quits on that day.

---

The timed_image_capture.sh simplifies usage of the python script on the pi. This script writes the Pi's captured images to /home/pi/trainingImages and pipes the output of the python script to /home/pi/trainingDebug for debugging purposes.

The recommended way to generate training images is by setting up a cronjob that launches the bash script on a daily basis.

---

Note: One may also launch the bash script manually, but it is important to note that previous images will be overwritten for that day, unless the numUploaded variable is manually increased. For example, if the current folder contains captures up to cap123.jpg, one should increase numUploaded to be 124, so that the previous images are not overwritten.

Note: it may be necessary to modify the path to the timed_capture.py script. The bash script currently assumes that the user's script is located in /home/pi/data_gathering/timed_capture.py, this may be changed if necessary.