Image classification and object detection have greatly improved over the last few years. It in now possible to use these technologies in a variety of use cases with very consistent results.
This repository contains files for development and deployment of a oilwell drilling bit dull grading system that uses computer vision to automatically grade bits and create reports.
Drill bits take quite a beating in downhole environment. Because of this there is quite a bit of damage seen on them which can give clues about operating parameters, future developments and design improvements. However, despite have a formal grading system it is not followed consistently resulting in large variations in grading and bad data collection.
This project aims to solve that problem. The app will be freely available to anyone to download and will also try to help improve the data collection process to make data more consitent and reliable.
The following modules are part of this repo
- Model files
- Notebooks for object detection
- Python scripts for testing various models
An example of semi trained model output
