Repository navigation
Project Vision and Scope
The commuters of Ngamwongwan Road who want to commute in the area soon and want to know how weather affects the traffic conditions.
Our product aims to help commuters predict the data of traffic flow and number of vehicles in a small section of the road, which will help commuters decide whether or not they need to change their plan or not.
Using images of vehicles on the road to analyze the relation between vehicle count per type, traffic flow, time, day of week, and weather. A pre-trained model was used to count the number of vehicles by type on the road, along with other metrics to predict patterns of road usage by time, day, and weather. This project aims to predict road usage at one part of Ngamwongwan Road, which runs passed Kasetsart University. The prediction of the data will be done using Gated Recurrent Units, Recurrent Neural Network model.
- Rust Rocket as a web server providing API
- Vue.js as a frontend framework
- Pytorch for training prediction model
- Using a Camera to collect pictures of cars in the street every 30 minutes interval for 3 weeks (using the pre-trained model to classify and count the number of vehicles on the street at a specific time)
- Pre-trained model used to classified: COCOv34
- Collect traffic flow data from Traffic Flow API by TomTom.
- Collect weather condition from Weather API by OpenWeather
Both data are collected within 1 minute after the image got captured and processed from the same location as the location that image got captured.
- Predict Number of vehicles on the road at specific periods and weather condition
- Predict the Number of each type of vehicles on the road at a specific period and weather condition
- Traffic flow and vehicle at specific time (Future) base on time and weather condition
- Descriptive Statistic of data used for model training (Flow, Weather, Vehicle Count)
- Simple Chart to visualize data
- Line Chart for Traffic Flow and Vehicle Counts over the time
- Pie Chart show number of datapoint on each weather condition
- Form to try with Prediction API
- Prediction for 5 hours after the current time (default on the site)
- Integrated Weather Forecast (Using API to obtain the data)
- Data is collected over short periods, which may also have noise, and missing data.
- The prediction model will train not to take the day of week into account, which may add inaccuracy to the result.
- Data are collected from just 50 meters portion of Ngamwongwan road therefore, data may be unable to represent the whole road situation.