Development process

Contents on development process such as Project Schedule, Trouble Shooting, Mentoring, ETC.

Project Schedule

2020년 날짜	내용
09.17	Final Selection of participants
09.17 ~ 10.06	Pre-preparation
09.18 ~ 09.23	Real-Time Mask Detection Code
10.06	IOT Mentor Application (김두훈 멘토 )
10.06 ~ 10.07	Github readme arrangement
10.07 ~ 10.08	Send messages with Kakao API
10.07 ~ 10.08	Extracting handwriting using tesseract-ocr
10.08 ~ 10.10	Using Google-Vision-API
10.08 ~ 10.11	GUI Environment
10.11 ~ 10.12	Send messages with telegram
10.18 ~ 10.19	Team Logo
10.19 ~ 10.23	Creating a thermal imaging camera code
10.21 ~ 10.22	Video Demo
10.24	First mentoring
10.24	Flow chart, diagram
10.29	Second mentoring
10.29~10.31	GitBook
10.24 ~ 10.31	PPT
10.31	Third mentoring
10.31	Hackathon End

Trouble Shooting

Tesseract

The name tag detection was attempted through the Tesseract-OCR before the Google Vision API, which is currently used for name tag detection. However, due to the lack of data model learning and the nature of the uniform name tag, problems were identified that were difficult to extract names even if images were pretrained for detection. Therefore, although the model has been selected for detection of name tags by using Google Vision API, which has a large amount of model training, we believed that if the identification of military uniform name tags is continuously trained, it will be possible to extract names with Tesseract-OCR without external connections.

KAKAO API

Before Telegram, which is currently used for sending notifications, we tried to send messages using Kakao API. However, in order to send a picture, which is one of the important data, an image URL was required, which means that communication between the Raspberry Pi and the server is necessary. This was judged to be the use of small swords to catch chickens, so we chose Telegram that can be sent by the Raspberry Pi itself and has excellent security. This problem is considered to be a problem that will naturally be solved if existing systems such as military short message transmission system are used when used in the future defense network.

Mentoring

First Mentoring

Mentoring was conducted through Google Meet with team members for a total of 2 hours and 20 minutes from 15:00 to 17:20 on October 24. The contents of the mentoring are as follows.

• License - Described the importance of copyright in the created content and all development content.

• Blueprint - Raspberry Pi blueprint need to be supplement

• Manual - Explain the importance of How to Build/How to Use

Second Mentoring

We were mentored for 30 minutes from 18:00 to 18:30 on October 29th.

• Video Demo - Hardware part reinforcement Required

• ReadMe - Create Hybrid and Organize Required Overview

Third Mentoring

We were mentored for 1 hour and 30 minutes from 15:00 to 16:30 on October 31.

• PPT - Need to complement readability and emphasize manual. When creating slides, need to supplement the story line and then create

• PPT Manual - Need to supplement details

ETC

Going Outside

In order to secure the test dataset, we tried to take as many pictures as possible while the camera was available. With cameras not available inside the military for security reasons, these Test Datasets were one of the most important materials in carrying out the project and were able to be used for Mask Detection testing and REAL-TIME implementation.

Dawn Study

The project can always start around 17:30 p.m. due to the fact of online hackathon rather than training camp. In addition, it was possible to invest about two hours, excluding essential personal maintenance hours such as meals and showers, and to supplement this, night-time annual leave was implemented. If study was not possible due to work, the project could be carried out through efforts such as replacing the shift with another time zone.