bdd-driveratt

Project in collaboration with Berkeley Deep Drive on driver's attention in manual, semi-autonomous and autonomous vehicles. It uses Pupil Labs and a driving simulator.

The main goal of this code is to allow for collection and analysis of eye movement data from Pupil Labs glasses using their API.

This project includes:

• Makefile for all requirements (in progress!).
• Recording: Matlab and Python scripts to record eye movement data using Pupil Labs glasses.
• Preprocessing: Preprocessing pipeline for Pupil Labs glasses.
• Eyetracking_analysis:
  • Analysis code for categorization of eye movements during driving. Based on Tensorflow models.
  • Analysis code for comparison of gaze distributions of different conditions in driving.

Teresa Canas-Bajo

Instructions for Use of Eye-tracking Code

1. Go to the relevant folder (recording, preprocessing or analysis) and follow the instructions there.
2. The folder lib has three submodules needed for this code:

• pupil: Pupil labs API, under 'pupil': https://github.com/pupil-labs/pupil (commit: ec6171834129189d6e79eacc46ee2c39eb4d97e1)
• Pyuvc: Python bindings for the Pupil Labs fork of libuvc with super fast jpeg decompression using libjpegturbo (utilizing the tubojpeg api): https://github.com/pupil-labs/pyuvc
• nslr_hmm: Python implementation of the NSLR-HMM eye movement identification algorithm: https://github.com/pupil-labs/nslr-hmm

To do:

• Add simulator environment scenarios (coded on Prescan/simulink) for driving experiments.