Machine Learning Intern @ Hyperworks Imaging

01 March 2024

[Advancing Automated Well Log Digitization]



From March 2024 to July 2024, I worked as a Machine Learning Intern at Hyperworks Imaging in Bengaluru, India, where I focused on developing automated systems for digitizing raster well log images. My contributions significantly enhanced efficiency and accuracy in information extraction for machine learning applications.

During my tenure, I implemented a Mask R-CNN model from the Detectron2 library to successfully extract multiple segments from well log images. This approach achieved an 80% precision rate and stored the extracted results in JSON format for further processing.

Additionally, I engineered an innovative solution leveraging Generative Adversarial Networks (GANs) to remove gridlines and a transformer model to map curves, enabling precise digital representation of well log data. Furthermore, I designed and executed an OCR-based system that incorporated clustering and post-processing techniques to accurately extract depth information from well logs.

To ensure high accuracy and reliability, I validated the digitization results against corresponding LAS files, facilitating seamless integration for downstream analysis.