Rebecca Li

Rebecca Li

Machine Learning Engineering Manager

A cross-library image augmentation module for Deep Learning Training

Developed a versatile image augmentation framework incorporating 300+ operations from 8 libraries, enabling seamless cross-library functionality, advanced transformation pipelines, differentiable checks, and dynamic customizability for users.

Keywords: Image Augmentation, Transformation, Cross-library, Imgaug, Albumnetation, torchvision

  • Adobe Research Opensource Project, hands on 80% code contribution
  • Main module building: Rebecca Li, Yannick Hold-Geoffroy, Geoffrey Oxholm
  • Website: https://adobe-research.github.io/beacon-aug/
  • Code: [Python]
  • Award: Top Session Award at Adobe Tech Summit 2022 (together with the whole Beacon module)

Li, X. R., Hold-Geoffroy, Y., Oxholm, G., Singh, K. K., Zhang, Z., Zhang, R., Andriushchenko, M., & others. (2021). Beacon-aug: A cross-library image augmentation toolbox. GitHub. Retrieved January 11, 2022, from https://github.com/adobe-research/beacon-aug