Using Artificial Intelligence to Detect-Distracted Driving
- 0 Collaborators
This project was made for TKS AI Hackathon in October 2020. Focused on solving distracted driving accidents (causes >96% of all 1.35 million deaths) by leveraging artificial intelligence algorithms such as transfer learning + Convolutional Neural Networks ...learn more
Project status: Published/In Market
Intel Technologies
Intel CPU
Overview / Usage
This project was made for TKS AI Hackathon in October 2020. Our goal was to identify a problem that can be solved using Artifical Intelligence. The problem that our team has come up with is the problem of distrcated driving. More than 50 million people are caught up in car accidents every year. What if there was a way we could prevent it? Self driving cars aren't feasiable; they can't be implemented right now. We need to find a way to implement a method to detect distracted driving in our cars.
Methodology / Approach
We use a Convolutional Neural Network to detect distracted driving in the dataset. Our neural network has 5 layers (1 input, 1 output, 3 hidden layers) with each hidden layer following this process:
- Convolution
- ReLU
- Maxpooling
- ReLU
- Dropout
This format is used for 3 layers. The kernal size for the convolution is kept the same (3x3, no padding) along with maxpooling (2x2, no padding). Dropout is kept at the same value for all 3 layers (0.1). The output activation at the end is softmax resulting in a probability distribution for each class.
More info found at: https://github.com/srianumakonda/Using-Artifical-Intelligence-to-Detect-Distracted-Driving
Technologies Used
Tensorflow keras, Numpy, PIL, Python, Pandas, Matplotlib, OS, datetime, pytz
Repository
https://github.com/srianumakonda/Using-Artifical-Intelligence-to-Detect-Distracted-Driving