/MVA-RecVis-a3

An image classifier of birds

Primary LanguagePython

Bird images classification

Requirements

Run the following command to install required dependencies:

pip install -r requirements.txt

Dataset

The dataset is taken from the CUB-200-2011 dataset.

The training/validation/test images can be downloaded here.

For some experiments, the birds need to be extracted from the image: execute the instructions in colab_extract_birds.ipynb.

How to obtain the results?

These are the commands for training for each experiment:

  • (2.1): python train.py --exp dn161-1 --arch Densenet161 --data bird_dataset --no-train-data-aug --no-val-data-aug
  • (2.1)+(2.2): python train.py --exp dn161-12 --arch Densenet161 --no-train-data-aug --no-val-data-aug
  • (2.1)+(2.2)+(2.3): python train.py --exp dn161-123 --arch Densenet161 --no-val-data-aug
  • (2.1)+(2.2)+(2.3)+(2.4): python train.py --exp dn161-1234 --arch Densenet161

Then to evaluate:

python evaluate.py --exp dn161-1

You can also use the colab_classify.ipynb in the repo.

Acknowledgments

Adapted from Rob Fergus and Soumith Chintala https://github.com/soumith/traffic-sign-detection-homework.