Team Name
TSD Makers
Timeline
Summer 2021 – Fall 2021
Students
- Unique Karanjit
- Anish Thakur
- Sujan Shrestha
- Madan Bhurtel
- Darpan Dhimal
- Shimon Kumar Bhandari
Abstract
TSD is an android application that alerts drivers about traffic signs and signals that they encounter while driving to prevent road accidents and hazards. TSD uses the phone camera and the live video feed from the camera is passed into the application, then the trained machine learning model looks for traffic signs in the feed. Once traffic signs are detected and classified, a notification with beeping sounds pops up and warns the user. Users can also choose to turn off those alerts for a certain location
Background
Distraction and human error while driving lead to failure in reading the traffic lights and sign on the road. There might be an accident when one fails to read signs or traffic lights while driving. This problem could be solved if drivers were to be warned about the traffic signs and lights. This prevents drivers and pedestrians from accidents, property damages and leads to safety.
To warn drivers about the traffic signs and lights, we created an Android application that warns and notifies about every traffic sign and traffic light on the way the user is driving along with the map recognition of neighborhood roads and the signs and not notifying on those cases. The TSD application will receive the traffic sign and traffic light image information which is then recognized and warned to the user.
Project Requirements
Our application TSD has several customer requirements, as intended by the users. The project focuses on the functionality of detecting signs which are notified in our application. There are several features encountered by the users and they are provided with the option of having several features. The features are required by customers to maximize the attentiveness and liberation of using the application. The customer requirements explain how exactly the application works and what customers can access through our apps.
- DETECT TRAFFIC LIGHTS
- ABILITY TO LOCATE THE USER AREA
- NOTIFICATION OFF BASED ON LOCATION
- APPLICATION SHALL START IN LESS THAN 5 SECONDS
- APPLICATION SHALL MUTE THE NOTIFICATION FOR USERS DESIRED LOCATION
System Overview
We have built an android application for detecting traffic signs and traffic lights. This app receives frames from the external camera and processes that image to detect signs. After a sign is detected, it warns the user about those signs in the form of notifications. The notifications are provided only for the signals that the user asked for. We have the external camera as the device which captures the images of the environment from the dashboard. The camera turns on when the car is started and then connected with the phone through the cable. The video is streamed into the application. The application will look for the traffic signs in the video for those signs for which the user has chosen to receive notifications. The data is fed into the application which further uses a trained model to classify the images. The image classification is carried out with the trained and validated models from the dataset that holds all the signs.
If the traffic signs are detected in the video, we look at that in a dataset and if the detected object is a traffic sign for which the user wants to get the notification, our application notifies the user about the traffic signs simultaneously and consecutively. Along with that, we have another dataset that includes the home area location. We use Google Map API to know if the location is a home area. If the user does not want the notification about the signs in the home
area they won’t be notified. Every notification that the user wants to be notified is displayed in the user GUI.
Results
Demo video : https://youtu.be/BQDMl6_DUqs
Future Work
We will be adding Traffic light detection to the model and expanding app distribution to iOS as well.
Project Files
Project Charter (link)
System Requirements Specification (link)
Architectural Design Specification (link)
Detailed Design Specification (link)
Poster (link)
Source code
GitHub Repository: https://github.com/realAnish9/TrafficSignDetector-Final
