This is a selenium dynamic web page scraper to extract image URLs and download them to the local machine. This was originally intended to collect training data for the ZotBins waste image recognition project, but could be used for other purposes.
The script will download up to 400 images for a given search term. Runtime can vary depending on physical machine power and internet latency. During testing, the script would complete in 3-4 minutes for a single search term.

• Python: (tested on version 3.6.5)
• Mozilla Firefox (tested on version 72.0.2)
• Setup selenium: https://selenium-python.readthedocs.io/installation.html
• only tested on Windows 10

python download_images.py <search terms>

For example, to download images for utensils and water bottles, run:

python download_images.py utensils "water bottles"

Remember to enclose multi-word arguments in quotes.

Web scraping is by nature dependent on the target. Therefore, any updates on google images may invalidate the script. View the download_images.py script documentation to see what can be done to keep the script updated.

During testing, we encountered the following error a few times:

selenium.common.exceptions.ElementNotInteractableException: Message: Element <option> could not be scrolled into view.

Somehow, we recieved the error the first one or two times we ran the scripts, and running the script afterwards did not reproduce the error.
We suggest that users who run into this error just run the script multiple times, and hopefully the error will disappear. The cause of this error is still unknown...