mehmetkeremturkcan/cs3-ai-teaching-kit

CS3 AI Teaching Kit

38 lessons for the Jetson teaching kit, ready to teach and ready to run. Open index.html for the lesson list.

There is also an advanced course/ with 10 ready-to-run programs (scripts only, no decks) that combine the camera and sensor courses — see its README.

This offline package is one component of the hybrid CS3 AI Teaching Kit developed for the 2026 CS3 Research Experience for Teachers program at Columbia University. The companion online AI Training Toolkit contains 17 browser lessons.

Download to the Ubuntu Desktop

Open a terminal on the Jetson or another Ubuntu computer and run:

sudo apt update
sudo apt install -y python3-venv

python3 -m venv "$HOME/.venvs/huggingface"
"$HOME/.venvs/huggingface/bin/pip" install --upgrade huggingface_hub

"$HOME/.venvs/huggingface/bin/hf" download \
  Center-for-Smart-Streetscapes-CS3/cs3-ai-teaching-kit \
  --repo-type dataset \
  --local-dir "$HOME/Desktop/cs3-teaching-kit" \
  --force-download

Use the same command for later updates. --force-download overwrites packaged files with the newest published versions. Local files that are not part of the dataset remain in place. Close open lesson programs before updating.

After the download, open:

~/Desktop/cs3-teaching-kit/index.html

What a lesson folder holds

File	Purpose
<Lesson>.html	slide deck; open in a browser and present
<Lesson>.pdf	the same 781 slides as pages, for printing or handing out
<Lesson>.md	the note the deck was built from
*.py	the programs the lesson teaches
media/	the figures used by the note and the deck

Model and data files (*.caffemodel, *.onnx, *.xml, *.engine, car.mp4) sit beside the programs that load them, so a lesson runs from its own folder.

Presenting

Open the .html deck in any browser. It carries its figures inside itself, so it needs no network and no other files.

Key	Action
Right arrow, Page Down, Space	next slide
Left arrow, Page Up, Backspace	previous slide
Home, End	first slide, last slide
f	full screen

Running a lesson

Move into the lesson folder first, because each program loads its models and data by relative path:

cd ~/jetson_Course/sensor\ course/1_LED
python3 1_LED.py

Press Ctrl+C to stop a sensor program; press Esc or q in the video window to stop a camera program.

Vehicle recognition, two versions

Lesson 13 of the camera course ships both (i) 13_Car_recognition.py, the PyTorch model, and (ii) 13_Car_recognition_tensorrt.py, a TensorRT engine built for this board, which runs about three times faster. The prebuilt yolo26n.engine is included; export_tensorrt.py rebuilds it if a board ever needs its own.