/BirdCLEF_2024

Codebase for BirdClef 2023 solution

Primary LanguagePython

Bird Clef 2023 1st place solution

Codebase for Bird Clef 2023 1st place solution

Pre-requirements

  • Ubuntu 22.04.2 LTS
  • CUDA Version: 12.0
  • CUDA Driver Version: 525.105.17
  • conda 22.9.0
  • GPU A100 40Gb (but any other GPU, compatible with CUDA Version and with VRAM >= 10Gb, should work)
  • Hard Disk: 1.5 Tb
  • RAM: 500 Gb

Main Pipeline

# Export credentials
 export KAGGLE_USERNAME={KAGGLE_USERNAME}
 export KAGGLE_KEY={KAGGLE_KEY}
# Create environment
conda env create -f requirements/environment.yml
conda activate bird_clef_2023_1st_place
pip install -e .
# Run Pipeline
bash rock_that_bird.sh "{GPU_TO_USE}" # By default: bash rock_that_bird.sh "0"

Find final model in logdirs/convnext_small_fb_in22k_ft_in1k_384__convnextv2_tiny_fcmae_ft_in22k_in1k_384__eca_nfnet_l0_noval_v32_075Clipwise025TimeMax_GausMean/onnx_ensem/model_simpl.onnx