Skynet is trying to develop machine learning solutions to solve the problems faced in design field of construction industry. A machine learning algorithm can solve the design problems faced while collaborating within multi-disciplinary design teams. A machine learning algorithm thus will reduce the time spent on redesigning and cost for the same. It will also help with better collaboration within teams by decreasing the number of conflicts between the multidisciplinary design teams.

A study from Engineers Daily estimated that design errors accounted for 38% of construction disputes. In terms of money percentage, design errors, directly and indirectly, constitute almost 14% of total contract money. One of the main reasons for design errors is less collaboration between different design teams working on a single project.

60% of time is wasted on redesign and solving clashes

A machine learning algorithm which takes input from all departments and optimises the design of ceiling to suit the requirements from all departments

The task is divided into 3 phases- identify the components in the drawings provided, learn all the requirements from all design departments and optimise the design of ceiling to meet the design requirements.

1. Identify components in the drawing
2. Learn all requirements from departments
3. Optimise the design of ceiling to meet the requirements

In the first stage, neural network was used to detect the object in drawings. Initially, limited data set was given to train the neural network.

The method of transfer learning was used in which existing codes applied to different context are modified to suit the needs of this project. A Tensorflow object detection- faster RCNN (Regions with Convolution Neural Network) inception-v2-COCO model was used as a base to develop from.

Team- Grevil Colaco, Kishan Basavraj, Somanath Pal and Animesh Mazire

The project is ongoing, please get in touch for more information about the project

Get in touch

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.