├── README.md
├── requirements.txt
├── src
│ ├── __init__.py
│ ├── config.py
│ ├── dataset.py
│ ├── metric.py
│ ├── network.py
│ └── util.py
├── test.py
└── train.py- Create python virtual environment
$ python3 -m venv venv/aue8088
$ source venv/aue8088/bin/activate- Check whether the virtual environment set properly
: The result should end with
venv/aue8088/bin/python.
$ which python- Install required packages
$ pip install -r requirements.txt- Login
$ wandb login- Specify your Wandb entity
$ echo "export WANDB_ENTITY={YOUR_WANDB_ENTITY}" >> ~/.bashrc
$ source ~/.bashrc- Please complete this function to measure accuracy of the prediction
- Please write MyF1Score class from scratch
- Calculate per-class F-1 score in a one-vs-rest manner
- Apply this new metric (hint: update src/network.py)
- Try different settings (src/config.py)
$ python train.py- How to improve performance more?
- Find state-of-the-art method/model(paper) on TinyImageNet-200 dataset
- Check difference between baseline and state-of-the-art
- Apply missing stuff in the baseline