Leading IT Company envisioned to Automate Data Processing from blueprints for Electrical Engineers

The client is a global professional services firm focused on delivering digital transformation for their clients, putting digital and data to work to create a competitive advantage.

The client wanted to create a solution for electrical engineers, enabling them to derive meaningful data from electrical blueprints to enable informed cost decision-making.

The team, after understanding the client’s process, architected a solution that would use Computer Vision and Machine learning to extract data. This data was further processed to provide actionable information

Convert blueprints into images for further processing in computer vision
Provide data processing based on predefined rules
Ensure deep learning to make the tool more intelligent with time

Through its innovative solution, the client wanted to equip electrical engineers with meta-information on their blueprints at the click of a button. Existing solutions were limited to giving data using OCR tools. Such solutions gave

Limited Use of error-prone data. Also, the output data was dependent on input data quality and was, therefore, unreliable.
In-efficient process OCR tools only provided image capture and not data capture. Data still had to be typed in using manual, error-prone methods.

By itself, typed-in data with image capture could not provide any actionable information to the engineers.

The team determined the steps required to provide actionable information from blueprints. Trantor chose Python as the software language. Keeping the blueprint in mind, they chose Computer vision and Machine learning in the backend to convert blueprints into Images and images into data. This data was further processed into structure format by using predefined rule sets.

Blueprints were converted into Images for Computer vision and Machine Learning(ML).
The processing flow was tailored to the characteristics of the source material
Using Computer Vision, the regions were detected in which the relevant data was present.
Machine learning was used for data extraction from detected regions.
Data were processed using built-in business rule sets
Processed data was moved into the server for easy retrieval and search;
Intuitive, user-friendly UI was designed and built to allow end-users to convert blueprints into actionable data in a few clicks

30% improvement in bottom-line due to informed decision-making
60% reduction in operational cost
95% data accuracy

=Tech Stack Used:

Python
ROR
Open CV
ML (KERAS)
Tesseract APIs