CARLA capture → Mininet-WiFi encrypted replay → packet grouping → frame-size features → vehicle visibility → anonymous passage events.

Use a separate Python environment compatible with the pinned dependencies and your CARLA Python API. The NumPy pin requires Python below 3.11. Install CARLA and Mininet-WiFi separately, along with ffmpeg, ffprobe, tshark, mergecap, capinfos, and tcpdump. Mininet-WiFi requires a Linux networking environment and root privileges for emulation.

From this folder:

python -m pip install -r requirements.txt

The original scripts contain machine-specific dataset and checkpoint paths, including /media/ubuntu/research/ and /home/ubuntu/carla. Update the relevant script settings before running. Raw captures, trained weights, and normalization files are not bundled.

Follow the detailed guide for the ordered capture, training, inference, and export commands.

The final export groups passage events by capture run. Part 2 evaluates tracking using the measured detector-error profile; its default experiments generate their own trajectories and events.