ailuropodaWu / FGVC-classifier

FGVC Classifier on CUB-200-2011

FGVC Classifier on CUB-200-2011

NYCU Intro. ML final project

Performance

• Accuracy 85.9% on both public and private leaderboards
• Leaderboard on Kaggle contest

Training Detail

• Training environment
  • Python version: 3.10.13
  • framework: Pytorch
  • Hardware: NVIDIA 1080Ti with CUDA 12.3 & Intel i7-12700
• Pretrained model
  google/vit-base-patch16-224
• Hyper-parameters
  • epoch: 5
  • batch size: 10
  • learning rate: 2e-5
  • weight decay: 1e-4