Imageflipper is a program that searches for images using google search and then displays these images continually. I built this program to run on a Raspberry Pi with a small LCD screen. The Pi is mounted on a wall in my study and it brings a continual source of new and exciting images to my eyeballs. In a way, you can think of it as a smart screen saver.

• Display images based on configured search terms
• View, retrieve, and delete images using the keyboard
• Downloads new images periodically in the background and deletes old images as well
• Blacklists bad urls based on failed image downloads
• Threads are used so new images are downloaded and old images a cleaned, without disrupting image display
• Logs requests, errors, etc
• A daemon also exists for ...
  • Adding new search terms
  • Clearing out old images
  • Downloading new images
  • Listing images that are currently being displayed
  • Showing images that have been downloaded and their disk utilization

git clone https://github.com/fjohnson/imageflipper.git
cd imageflipper
pip install pipenv
pipenv shell
pipenv install
*Now edit your config.json file*
python display.py

This program uses Google's custom search API. You will need two pieces of information, an API key and a search engine ID, which you can get here: https://developers.google.com/custom-search/v1/introduction

Next, fill in the config.json file. Here's an example.

{
"image_download_interval": 36000,
"flip_frequency": 5,
"search_terms": ["garfield"],
"extra_images": ["pizza","cheese"],
"user_agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/70.0.3538.110 Safari/537.36"
"results_per_page": 3
"search_engine_id": '...'
"api_key": '...'
}

And here's what the variables mean.

image_download_interval: How often google is queried for new images in seconds
flip_frequency: How long to wait until the next image is shown
search_terms: The terms you want to find images for
extra_images: If you previously downloaded images or you just have images lying around, you can add them to be displayed. Any images in the images/pizza and images/cheese dir will be added in this example.
results_per_page: How many urls are returned in a search? The max is 10.
user_agent: Enter whatever user agent to masquerade here as.
api_key: Your api key
search_engine_id: Your search engine id

Start it up with python display.py

Defined inputs are:
q quits
d force a download of new images
left arrow: go back an image
right arrow: go forward an image
i: show image position out of total
delete: delete an image from the display buffer and disk

The daemon is started when display.py is run and by default accepts TCP connections on port 9999. Edit SearchTermServer.py to change the port.

The daemon accepts "raw" tcp connections. No protocol here; open up a tcp connection and start typing. You can accomplish this with PuTTY, telnet, netcat, etc. Once in type ^commands for the list of options.

• Implement an image comparison algorithm and delete duplicate images.

  Note: I tested out image comparison using the "Structural Similarity Index" method (see here) but it was too slow. Comparing a newly downloaded image requires a check against every previously downloaded image. If you have hundreds of images, it can be done on the PI, but it can take literally hours. I decided it wasn't worth it in the end. Another option could be a standalone program that runs in the background and cleans up duplicates.

• Display images based on a dictionary of random words instead of predefined keywords.