-
Prepare the Dataset:
- Follow the instructions in the SVDD Challenge CtrSVDD_Utils Repository to manage the dataset.
-
Organize Dataset Structure:
- Ensure your main directory has the following structure:
Datasets/ ├── dev/ ├── train/ ├── eval/ ├── dev.txt └── eval.txt
- Ensure your main directory has the following structure:
-
Set Up the Environment:
- Create a conda environment using the provided
requirements.txt
file:conda create --name your_env_name --file requirements.txt conda activate your_env_name
- Create a conda environment using the provided
-
Run Training:
- Execute the training script by specifying the base directory of the dataset:
python train.py --base_dir {path_to_Datasets_folder}
- Additional arguments can be added, such as
--algo
for the rawboost algorithm:python train.py --base_dir {path_to_Datasets_folder} --algo {algorithm_choice}
- To change the model, modify the model selection directly in the
train.py
script header.
- Execute the training script by specifying the base directory of the dataset:
-
Run Evaluation:
- Execute the evaluation script by specifying the base directory of the dataset:
python eval.py --base_dir {path_to_Datasets_folder}
- Execute the evaluation script by specifying the base directory of the dataset:
-
Custom Arguments:
- You can customize various parameters through command-line arguments as needed.
- Example:
python train.py --base_dir {path_to_Datasets_folder} --batch_size 64 --epochs 50
-
Changing the Model:
- To use a different model, edit the model import and instantiation in the
train.py
file.
- To use a different model, edit the model import and instantiation in the
For further details, refer to the code comments within the scripts.