https://github.com/sushmitha-93/distracted_driver_detection
To detect if a driver is distracted or not using State farm's distracted driver image dataset to train Resnet50 model
https://github.com/sushmitha-93/distracted_driver_detection
axios bootstrap5 convolutional-neural-networks docker fastapi javascript python reactjs resnet-50 tensorflow
Last synced: 3 months ago
JSON representation
To detect if a driver is distracted or not using State farm's distracted driver image dataset to train Resnet50 model
- Host: GitHub
- URL: https://github.com/sushmitha-93/distracted_driver_detection
- Owner: Sushmitha-93
- Created: 2022-06-10T01:36:09.000Z (about 4 years ago)
- Default Branch: main
- Last Pushed: 2023-07-04T02:07:17.000Z (almost 3 years ago)
- Last Synced: 2026-01-03T13:55:37.627Z (6 months ago)
- Topics: axios, bootstrap5, convolutional-neural-networks, docker, fastapi, javascript, python, reactjs, resnet-50, tensorflow
- Language: JavaScript
- Homepage: https://sushmitha-93.github.io/Distracted_Driver_Detection/
- Size: 4.01 MB
- Stars: 2
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
### Checkout Demo deployed at https://sushmitha-93.github.io/Distracted_Driver_Detection/phase1demo
Model deployed at: http://44.203.70.143:8500/docs
Distracted driving is causing about 3000 deaths and 280,000 injuries per year according to National Highway Transportation and Safety Administration (NHTSA). Unfortunately distracted driving has become all too common these days with advent of smart phones and social media.
Computer vision solutions based on machine learning image classification algorithms can be effectively used to detect inattentive drivers using any sort of dashboard cameras and alert drivers.
### Demo
Backend: FastAPI to take image as requests and respond with prediction result using trained Resnet ML model to make predictions. FastAPI Docker container image is deployed and hosted on Heroku.
Frontend: ReactJS application, Hosted on GitHub pages.
Check out Phase 1 demo which uses the trained Renet-50 model to predict unlabelled test images