Skip to content

Latest commit

 

History

History
14 lines (11 loc) · 454 Bytes

README.md

File metadata and controls

14 lines (11 loc) · 454 Bytes

birds_classification

Install

  1. conda create -n birds python=3.10
  2. conda activate birds
  3. python3 -m pip install -r requirements.txt
  4. Install apex as doc

Training process

  1. train.sh (may be resume.sh if the machine crash)
  2. finetune.sh (train with small categories)
  3. eval.sh (get 'incorrect' samples for the model)
  4. train_second.sh (train with 'incorrect' samples enhanced)