According to Times of India about 146,133 people were killed in road accidents in India in the year 2016. Unfortunately about 30% of deaths are caused due to delayed ambulance.
Every second heart attack patient in India takes more than 400 minutes to reach a hospital, which is almost 13 times more than the ideal window of 30 minutes, government data shows.A two-year data from the ongoing Management of Acute Coronary Event (MACE) Registry of the Indian Council of Medical Research (ICMR) shows at some places it even takes 900 minutes as a lot of time is wasted in transportation.
In case of fire department, such is the condition of traffic during peak hours in the city that a heavy fire truck rushing to tackle a fire breakout takes about four minutes to traverse just one kilometre. That was one of the findings by Telangana State Disaster Response and Fire Services department officials when they conducted empirical speed tests last month to know the time their vehicles take to reach destinations.A senior official said that the heavy fire truck, which can carry 4,500 litres of water, clocked only 15 kilometres per hour (kmph), while taking four minutes to travel one km on an average. The mini truck which has a capacity of 1,000 litres also showed the same result during the tests, which were conducted from the Secretariat fire station.
The case is no different for the police department also, moreover these statistics and studies are from India, but the problem with the emergency services is more or less the same across the globe.
There are numerous works by various people already on this issue, but the solutions like constructing separate roads/lanes for emergency vehicles or reducing the distance between their stationed area and place of response isn't as practical or possible in some cases as it sounds. So if there could be a way that these emergency vehicles could be prioritized, when on road could get a real decrease in their response time but the first step in doing so, is to recognize them among all other vehicles on busy road during rush hours, which is what this project is aimed at.
-
OpenCV is used for feature exraction with a keras model to predict upon that.(Reason for its slow speed)
-
OpenCv: Haar cascade is used for detection purposes
-
Deep Learning Model: Made on keras with tensorflow in backend
- Mode Architecture: A Functional API Model
- Transfer Learning without top (VGG16 Model)
- Dense Layer(s) --4
- Droplayer(s) --2
- BatchNormalization Layer(s) --2
- Optimizer: Adam (default)
-
Requirements:
- Tensorflow
- Keras
- Numpy
- Pandas
- Scikit Learn
- OpenCv
- Seaborn
- Matplotlib
- Jupyter Server
- Conda (optional)
- Git (optional)
-
Mode Deployment:
- Model Deployed using REST API via flask (main.py)
- preprocessing for deployement (precessing.py)
- post request (request.py)
Work of : @thatdanish
An open source project
This space will be updated regularly with more details about the project soon
NOTE: Anyone is more than welcome to contribute a change/upgrade/bugfix etc.