paper link: http://alisec-competition.oss-cn-shanghai.aliyuncs.com/competition_papers/20211201/rank10.pdf
This work is sponsored by Natural Science Foundation of China(62276242), CAAI-Huawei MindSpore Open Fund(CAAIXSJLJJ-2021-016B), Anhui Province Key Research and Development Program(202104a05020007), and USTC Research Funds of the Double First-Class Initiative(YD2350002001)”。
We use Mindspore to develop our algorithm.
The generated partial image candidate set
Experiment results
How to run
- Obtain all Cifar10 data from the official website to the folder CIFAR-10-FANs-PY;
- Create two empty folders. new_data and train_model;
- Run gen_corr_data.ipynb to generate corruption data;
- Run final_gen_data_iter to generate the first generation of the data;
- Run train.py to generate the first generation model;
- Generate the next generation of the data.
- Loop 5 ~ 6 until num = 6. In general, it takes 5 times to train the model to get the final submitted model, and 4 times to get the final submitted data.npy.