Machine Learning / AI / Computer Vision Intern – Comparative Study of Balancing Techniques for Image Classification with Imbalanced Datasets (6 months)
We are Schlumberger, the leading provider of technology and services to the energy industry. Throughout much of the oil and gas lifecycle in over 120 countries; we design, develop, and deliver technology and services that transforms how work is done.
We define the boundaries of the industry by unleashing our talented people’s energy. We’re looking for innovators to join our diverse community of colleagues and develop new solutions and push the limits of what’s possible. If you share our passion for discovery and want to find out what you could really do, then here is the place to do it.
Imbalanced datasets with long-tailed distributions are widely spread in real-world scenarios including datasets acquired with remote sensing technologies in various geological settings. When used in classification tasks they cause a bias in a decision boundary for machine learning and deep learning-based solutions, resulting in a smaller feature space for rare classes. In this internship project, we aim to conduct a comparative study of balancing techniques to address this challenge.
Computer vision datasets with ultrasonic, electromagnetic and microscope borehole images have been already collected internally and from publicly available sources and will be available within the scope of this internship. This will enable us to explore the application of different balancing techniques including off-line data manipulation methods (over-sampling / under-sampling, stratified splits), cost function weighting and especially recently proposed plug-and-play approaches (on-line rare class samples generation, on-line pseudo-labels generation, etc).
The main goal of this internship is to build a cumulative foundation of techniques devoted to enable vision-based deep learning models to efficiently discover structure within datasets with imbalanced data distribution and generalize well to infrequent classes. This has a wide application benefiting well integrity and well construction projects across Schlumberger.
Schlumberger is an equal employment opportunity employer. Qualified applicants are considered without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, age, disability, or other characteristics protected by law.