/bird_species_classification

Primary LanguageJupyter Notebook

Bird Species Classification

Introduction

  • The model training is supposed to run inside a docker container environment
  • An extensive amount of speed-up is possible with a designated graphics card
  • You need a supported NVIDIA GPU for multithreaded training

To start the container, run:

docker-compose up --build

To free space of old containers, run:

docker system prune --volumes

How to use the code:

Important Before you get started, make sure you download the images of the dataset and place the min the corresponding target folder. (tf/input) The dataset is way too large for GitHub.

There are several files in this project serving different purposes:

  • training_default.py -> Execute for training without image pre-processing and augmentation
  • training_augmented.py -> Execute for training with image pre-processing and augmentation
  • logs_to_plot.py -> Creates plots based off model training logs
  • model_evaluation.py -> Evaluates test accuracy and loss on a model
  • clustering.ipynb -> Interactive clustering jupyter notebook