Border patrol app

https://border-app.azurewebsites.net

Table of Contents

• General Information
• Used Dataset
• Used Technologies
• Development process
• Architecture Diagram
• How does it work
• Demo
• Screenshots
• Team

---

General Information

Border patrol app is a web application which enables users to upload photos from cameras and then checks if there are people located on these photos. Users need to pass the authorization process and log in to the app to be able to use it. By doing this project we wanted to address a very recent problem of authorized crossing borders. It could be used as a system helping guards on borders to keep track of the eare around the border. It can be especially useful at night when there is a worse visibility for a human eye.

Tha app was created as a project no. 3 for a class "Introduction to applications and solutions based on Artificial Intelligence". The theme of a project was to create an app which brings some innovation to this world and has a potential of bringing benefits for people.

---

Used Dataset

Thermal Dogs and People Object Detection Dataset Dataset has examples of night and termal photos taken with border violators, as well as dogs.

---

Used Technologies

• Custom Vision Service
• Azure Storage
• Azure Web Services
• React.js 17.0.2
• Azure Logic App
• Azure SQL Database

---

Development process

• Finding a proper dataset of camera images made at night on the borders.
• Creation of detection system using one of Azure Cognitive Services - Custom Vision.
• Implementation of a frontend application using React.js.
• Connecting trained model with a frontend application using REST API and Azure Storage.
• Deployment of application using Azure Web Services.
• Design Logic Diagram.
• Connection of the app with Azure Logic App.

---

How does it work

1. A user opens the web application (hosted in Azure App Service) and is asked to log in to his Outlook account.
2. A user uploads a camera image and clicks the Upload button.
3. The web application sends the image to Azure Blob Storage.
4. Azure Blob Storage returns a url of the uploaded resource.
5. The web application sends the image url to Custom Vision API.
6. Custom Vision API returns predictions (detections) of people on the photo.
7. The web application shows a photo to a user with located people and probabilities.
8. User can adjust a threshold to make detections more or less accurate.
9. If threshold is above particular level system send mail to border controllers.

---

Demo

See a demo (in polish):

Demo

---

Screenshots

• Detekcja człowieka:

• Detekcja psa:

---

Team

Franciszek Wysocki

Krzysztof Kowalski

Patryk Peszko

Khameta Maksim

Selivonets Aliaksei

Kharashun Ilya

Podlesnyi Vladyslav