iNaturalist 2019

This project is part of a series of projects for the course Selected Topics in Visual Recognition using Deep Learning that I attended during my exchange program at National Chiao Tung University (Taiwan). See task.pdf for the details of the assignment. See report.pdf for the report containing the representation and the analysis of the produced results.

The purpose of this project is to implement a classifier for the iNaturalist 2019 Challenge. The implementation is based on the official PyTorch implementation of the paper "BBN: Bilateral-Branch Network with Cumulative Learning for Long-Tailed Visual Recognition".

1. Requirements

• PyTorch ≥ 1.0
• torchvision ≥ 0.2.2_post3
• TensorboardX
• Python 3.x

2. Dataset

The images and annotations can be downloaded from iNaturalist 2019.

3. Data format

The annotation of a dataset is a dictionary consisting of two fields: annotations and num_classes. The field annotations is a list of dict with image_id, fpath, im_height, im_width and category_id.

Here is an example.

{
    'annotations': [
                    {
                        'image_id': 1,
                        'fpath': '/home/BBN/iNat19/images/train_val2019/Plantae/7477/3b60c9486db1d2ee875f11a669fbde4a.jpg',
                        'im_height': 600,
                        'im_width': 800,
                        'category_id': 7477
                    },
                    ...
                   ]
    'num_classes': 8142
}

You can use the following code to convert from the original format of iNaturalist.

python tools/convert_from_iNat.py --file train2019.json --root iNat19/images --sp jsons

4. Pretrain Model

• BBN pretrain model for iNaturalist 2018

5. Usage

To train long-tailed iNaturalist2019 with imbalanced ratio of 50:

python main/train.py  --cfg configs/iNaturalist2019.yaml     

To validate with the best model:

python main/valid.py  --cfg configs/iNaturalist2019.yaml

You can change the experimental setting by simply modifying the parameter in the yaml file.

6. Credits

The Megvii-Nanjing GitHub Repository has deeply helped the development of this project.

7. License

Copyright (C) 2021 Alessandro Saviolo

This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program.  If not, see <http://www.gnu.org/licenses/>.