/CV_CIFAR10

CV_assignment3

Primary LanguageJupyter Notebook

Goal

  • train Cifar-10
  • follow constraints
  • maximize accuracy

Code Constraints

  • Maximum 11 layer
    • Layers that don't have parameter are not considered as layer.
  • Maximum 50 epoch
  • No extra DB. Use only Cifar-10
  • NO pretrain model. train from scratch

Skeleton Code

  • Followed VGG model : 78.31%
    • 8 Covn, 1 FC
    • 20 epoch

Report

  • font size: 10
  • 1 page
  • Itemize with contribution
    • When I change this, 5% accuracy increased.
    • Reason is this.
  • Include loss graph