# Script for fine-tuning our FSFM-3C model for MCIO LOO evaluation of cross-domain face anti-spoofing:

==============OCI2M==============
```
CUDA_VISIBLE_DEVICES=0 python train_vit.py \
    --pt_model ../../pretrain/checkpoint/pretrained_models/VF2_ViT-B/checkpoint-400.pth \
    --op_dir ./checkpoint/finetuned_models/MCIO_protocol/OCI2M/ \
    --report_logger_path ./checkpoint/finetuned_models/MCIO_protocol/OCI2M/performance.csv \
    --config M \
```

==============OMI2C==============
```
CUDA_VISIBLE_DEVICES=0 python train_vit.py \
    --pt_model ../../pretrain/checkpoint/pretrained_models/VF2_ViT-B/checkpoint-400.pth \
    --op_dir ./checkpoint/finetuned_models/MCIO_protocol/OMI2C/ \
    --report_logger_path ./checkpoint/finetuned_models/MCIO_protocol/OMI2C/performance.csv \
    --config C \
```

==============OCM2I==============
```
CUDA_VISIBLE_DEVICES=3 python train_vit.py \
   --pt_model ../../pretrain/checkpoint/pretrained_models/VF2_ViT-B/checkpoint-400.pth \
    --op_dir ./checkpoint/finetuned_models/MCIO_protocol/OCM2I/ \
    --report_logger_path ./checkpoint/finetuned_models/MCIO_protocol/OCM2I/performance.csv \
   --config I \
```

==============ICM2O==============
```
CUDA_VISIBLE_DEVICES=3 python train_vit.py \
   --pt_model ../../pretrain/checkpoint/pretrained_models/VF2_ViT-B/checkpoint-400.pth \
    --op_dir ./checkpoint/finetuned_models/MCIO_protocol/ICM2O/ \
    --report_logger_path ./checkpoint/finetuned_models/MCIO_protocol/ICM2O/performance.csv \
   --config O \
```