Detecting Distracted Drivers

The objective of this work is to successfully predict the likelihood of what a driver is doing in each of the pictures in the dataset¹.

The data consists on a set of images, each taken in a car where the driver is doing some action (e.g. texting, talking on the phone, doing their makeup). These are some examples:

The images are labeled following a set of 10 categories:

Class	Description
c0	Safe driving.
c1	Texting (right hand).
c2	Talking on the phone (right hand).
c3	Texting (left hand).
c4	Talking on the phone (left hand).
c5	Operating the radio.
c6	Drinking.
c7	Reaching behind.
c8	Hair and makeup.
c9	Talking to passenger(s).

Running the Code

Dependencies

• Python 3.6.1
• Tensorflow 1.3.0
• Keras 2.1.2
• matplotlib 2.0.2
• numpy 1.12.1

Command-line Execution

• Simple CNN in Keras

  Directory Path: /src/keras/base

  • Train the model: python train.py [-h] [--bsize BSIZE]

    Optional arguments:

    -h, --help	show help message and exit
    --bsize BSIZE	provide batch size for training (default: 40)

  • Test the model: python test.py [-h]

  • Predict from an image: predict.py [-h] [--image IMAGE] [--hide_img]

    Optional arguments:

    -h, --help	show help message and exit
    --image IMAGE	path to image
    --hide_img	do NOT display image on prediction termination

• CNN with VGG16 Transfered Learning

  Directory Path: /src/keras/vgg

  • Extract VGG16 deep features: python extract_vgg16_features.py [-h]

  • Train the model: python train_top.py [-h]

  • Test the model: python test.py [-h] [--acc] [--cm] [--roc]

    Optional arguments:

    -h, --help	show help message and exit
    --acc	will calculate loss and accuracy
    --cm	will plot confusion matrix
    --roc	will plot roc curve

  • Predict from an image: predict.py [-h] [--image IMAGE] [--hide_img]

    Optional arguments:

    -h, --help	show help message and exit
    --image IMAGE	path to image
    --hide_img	do NOT display image on prediction termination

  Note: Since the notebooks may not all be fully updated yet, the best way to run these programs is using the python scripts.

---

1: This dataset is available on Kaggle, under the State Farm competition Distracted Driver Detection.