Betoniertagebuch

• 

  Concrete diary filled out

• 

  Welcome Screen

• 

  Technology stack

🔥 Inspiration

Every construction site dealing with concrete with a compressive strength ability class >C25/30 is legally required to write a concrete diary. These diaries are often printed out as pdfs and filled out manually.

At the same time, a large number of documents about the supply of concrete in the construction industry are photographed and filed without making use of the resulting data.

We believe construction industry processes can and should be digitized! The emerging data should be captured and made use of. Therefore we planned to digitize data from delivery notes with text recognition and leverage this data for an automated concrete diary.

💡 What it does

BETONVERWALTUNG We developed a comprehensive concrete management software in the form of a web app including an online journalling of concrete diaries empowered by AI prediction tools.

Our app manages to automatically read out the text from concrete delivery notes and enables machine-empowered data documentation. Further, we want to provide additional benefits and improved planning by complexity prediction. The Deep Learning-based model predicts construction duration based on given construction site properties like required steps and seasonal data.

🛠 How we built it

This project has been built with a lot of love❤️, motivation🔥, JavaScript🌐 , and Python🐍, using:

• React
• Flask
• Pytesseract
• OpenCV
• SQLite database
• Google Colab
• SciKit Learn

🚧 Challenges we ran into

Our team was early aware that developing a full-stack solution and a Deep Learning model in 48 hours without is going to be a serious challenge. To top it all, we had difficulties finding application areas on the first day and just decided what we will go for Saturday at noon. Additionally, data availability was challenging as the given API did not provide enough data to easily apply machine learning methods. We, therefore, had to improvise by applying a neural network classifier instead of a regressor.

And yet, our most important challenge was not having an understanding of the construction business in general. To overcome this, we had to push through challenges arising from misunderstandings, such as the delivered and ordered quantity in delivery notes. We are proud to have overcome all emerging challenges and are happy about what we accomplished.

🎯 Accomplishments that we're proud of

• Making our automatic text recognition that could read-in delivery notes
• Setting up a good looking frontend
• Building a well-connected and scalable backend
• Training a well-performing AI model

All of these milestones made us very proud because we are progressing towards something that could really advance the construction industry.

📚 What we learned

Coming from diverse study backgrounds in mechanical engineering, industrial engineering, and information systems, we had very individual learning experiences. For Danil, it was the first time working with JavaScript and he learned to use React on the go. Lars understood the working of our Machine Learning model and developed a good understanding of the construction industry. For Jan it was the first time setting up a fully functioning backend.

Taken as a whole, each one of us learned a lot!

🔮 What's next?

We are looking forward to the feedback we are going to get for our solution! Given that we bump into serious customer need from the construction industry, we can imagine developing the web app further and improve the neural network with additional data sources.

Built With

• colab
• flask
• javascript
• keras
• opencv
• pytesseract
• python
• react
• scikit-learn
• sqlite