This repository contains:
PROJECT_DIR/image_transformation/
Please follow the instruction below to install it and run the experiment demo.
PROJECT_DIR/extra_data/
)PROJECT_DIR/dataset/
)Install dependencies
# COCOAPI
cd $DIR/libs
git clone https://github.com/cocodataset/cocoapi.git
cd cocoapi/PythonAPI
python setup.py build_ext install
# By default it creates 300 sample sets
python meta_set/main.py
# Save as "PROJECT_DIR/learn/mnist_cnn.pt"
python learn/train.py
# Get "PROJECT_DIR/learn/accuracy_mnist.npy" file
python learn/many_test.py
# Get "PROJECT_DIR/FD/fd_mnist.npy" file
python FD/many_fd.py
# You will see linear_regression_train.png;
# then check if FD and Accuracy have a linear relationship;
# If so, it is all good, and please go back to step 1 and create 3000 sample sets.
python FD/linear_regression.py
# Please follow the instructions in the script to train the model
# Save as "PROJECT_DIR/FD/mnist_regnet.pt"
python FD/train_regnet.py
If you use the code in your research, please cite:
@inproceedings{deng2020labels,
author={Deng, Weijian and Zheng, Liang},
title = {Are Labels Always Necessary for Classifier Accuracy Evaluation?},
booktitle = {Proc. CVPR},
year = {2021},
}
MIT