透過深度學習演算法萃取出鳥類型態的高階抽象特徵，探討鳥類功能性特徵與環境(緯度、海拔)的關聯性

Extracting high-level abstract features of bird through deep learning algorithms to explore the correlation between functional traits of birds and environment (latitude, altitude)

• Masked Autoencoders Are Scalable Vision Learners
• Github
• Ｍodel, Training Ｎote in mae/

• Training dataset:
  • ebird
    • web scrape from ebird
    • select top 100 by user ratineg per species
    • total species:
    • total imgs:
• Finetune and linprobe dataset:
  • iNaturelist 2021
• Note
  • Ebird Data Processing Notes

• Note
  • Using Ebird data to explore the functional characteristics and morphological diversity of birds worldwide
• Notebook for visualiztion in : mae/visual_mae.ipynb
• Traning log Visualization on Wandb