Usage

• Run pretrain model:

1. You can download pretrain model at: Baidu disk

   put unzip folder into code/model/{scope_name}, see {scope_name} in code.config.py

2. Run pre-train test using command:

   python main.py --code.config --action test

   where code.config.py is config file including CNN architecture, dataset name, and so on.
   

   The CNN module pre-train output is a csv file, which format with:

   {f1,f2,f3...,predict_label,true_label,quality_factory}, where f1,f2... is CNN confidence of each camera model.

3. Post-train in experiment folder

---

• You can train your personal pretrain model:

1. Download Drensden dataset into /code/dataset

   python main.py --code.config --action download --name Dresden

2. Generate a tensorflow records file:

   python main.py --code.cofnig --action generator

3. For training process, running command:

   python main.py --conf code.config --action train

4. Making your personal tampering dataset:

   python main.py --conf code.config --action splicing