DeepExtract is comprised of a suite of ML and computer vision models that perform the following key functions:

• Recognize and extract measurements and scale inference

• Identify and flag errors and engineering code violations

• Counting objects such as trees, sinks, doors, windows, etc

• Identify and extract metadata fields such as drawing number, title, etc

Estimators, quantity surveyors and procurement managers painstakingly eyeball over 400 engineering drawings per project, in order to manually count each all kinds of items that will need to be purchased for the building. Such items include, trees, sinks, bathtubs, doors, windows, fireplaces etc. This labor intensive tedious process typically takes one week to ten days, is error prone and often leads to cost overrun if over-estimated or scheudle delays in case of shortfalls.

• Minimize errors in BoM generation before purchasing eliminating waste

• Generating accurate counts ahead of purchasing enables negotating power and better planning

• Avoid cost overruns due to last minute POs

• Civil Code checks and suggested fixes for compliance

• Design template and style violation detection for junior arhictects, designers, and structural engineers

• Checking for cyclical references across drawing disciplines

• Checking for contradictory grid lines across disciplines