Protests, TikTok, and more 🎥: analyzing images and videos with AI

Dataharvest 2025

Jonathan Soma

Knight Chair in Data Journalism, Columbia University

If you're interested in scraping TikTok, that's a different game! You might be interested in this passive scraping session.

Notebook links

Open up the notebooks on Google Colab! You can press the play buttons to make them work.

You will need to install ffmpeg.

My PDF library Natural PDF also has a lot of bits and pieces that work with images.

Name		Name	Last commit message	Last commit date
cars		cars
tattoos		tattoos
.gitignore		.gitignore
01-What color is the sky.ipynb		01-What color is the sky.ipynb
02-Debate screen time.ipynb		02-Debate screen time.ipynb
03-Regions of interest.ipynb		03-Regions of interest.ipynb
04-Car Tracking and Counting.ipynb		04-Car Tracking and Counting.ipynb
05-Semantic search.ipynb		05-Semantic search.ipynb
README.md		README.md
birds.jpg		birds.jpg
car.jpg		car.jpg
cars.zip		cars.zip
cat.png		cat.png
city.png		city.png
istockphoto-534232220-640_adpp_is.mp4		istockphoto-534232220-640_adpp_is.mp4
sky.jpg		sky.jpg
tattoo.png		tattoo.png
tattoos.zip		tattoos.zip