Streamlining Business Operations with Handwritten Text Detection through Computer Vision
ThinkCol develops web application with Machine Learning capabilities for Leading Provider of Integrated Printing Solutions
A globally-recognized full-service legal and financial communications provider looks for a solution to automate a key task in their operation. Numerous printed documents and pages with handwritten annotations are to be reviewed and identified by its employees prior to further action – in peak season, each officer has to go through around 15k pages per day, just to categorize all of them into pages that needs to be further reviewed and those need not.
ThinkCol built a web application to run machine learning models to classify each page of a document, with over 97% accuracy*. The curated results can be downloaded by the users for further processing. Users now easily leverage the machine learning model to streamline their tasks.
Throughout the process, ThinkCol worked extensively with the client and created the web application and AI models using Agile, UI/UX and MLOps methodology. The machine learning module was written in Python and deployed in AWS Sagemaker; The web application was written in JavaScript and deployed in AWS as well.
* Recall rate is 97% while F2 score is 94.8%
