Cornell Birdcall Identification

Source code of solution for Cornell Birdcall Identification competition.

Solution

Key points:

• Log-scaled mel-spectrograms
• CNN models
• Low public/private score :)

Quick setup and start

Requirements

• Nvidia drivers, CUDA >= 10.2, cuDNN >= 7
• Docker, nvidia-docker

The provided Dockerfile is supplied to build an image with CUDA support and cuDNN.

Preparations

• Clone the repo, build docker image.

  git clone https://github.com/lRomul/argus-birdsong.git
  cd argus-birdsong
  make build

• Download and extract dataset to data folder.

Run

• Run docker container

make

• Create folds split

python make_folds.py

• Train model

python train.py --experiment train_001