Research Intern at Seoul National University

  • Developed data refurbishing, a new approach in data augmentation to make Deep Learning model training faster.
  • Implemented Revamper, a data loading system realizing data refurbishing on top of PyTorch library with 2K+ lines of code.
  • Accelerated the training of computer vision models(VGG16, ResNet-18, MobileNet-V1, EfficientNet-B0) by 1.03× – 2.04× while maintaining comparable accuracy.

<Summer 2020 - Winter 2020>