Just example you might need to try around
pip install -r requirements.txt
python prepare_trigger_set.py
Training model without watermark:
python train.py --dataset cifar10 --max_epochs 60
Training model with watermark:
python train.py --wmtrain -wmt --dataset cifar10 --max_epochs 60
Evaluating original model accuracy:
python evaluate.py --model_path checkpoint/original_model.t7 --dataset cifar10 --test_db_path ./data
Evaluating watermarked model accuracy:
python evaluate.py --model_path checkpoint/watermarked_model.t7 --dataset cifar10 --test_db_path ./data
Evaluating watermark effectiveness
python evaluate.py --model_path checkpoint/watermarked_model.t7 --wm_path ./data/trigger_set/ --wm_lbl labels-cifar.txt
Fine-tuning the watermarked model:
python fine_tune.py --model_path checkpoint/watermarked_model.t7 --dataset cifar10 --train_db_path ./data --test_db_path ./data --epochs 5
Evaluating fine-tuned model for watermark effectiveness:
python evaluate.py --model_path checkpoint/fine_tuned_model.t7 --wm_path ./data/trigger_set/ --wm_lbl labels-cifar.txt
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent