SP-ASFDA

implementation for SP-ASFDA (Self-Partitioning for Active Source-Free Domain Adaptation)

Prerequisites:

• python == 3.12.2
• pytorch == 2.2.1
• torchvision == 0.17.1
• numpy, scipy, sklearn, PIL, argparse, tqdm

Dataset:

• Please manually download the datasets Office, Office-Home, VisDA-C from the official websites, and modify the path of images in each '.txt'. The source model of office-home and visda can be downloaded in this Url

Training:

Active Source free domain adaptation (ASFDA) on the Office/ Office-Home dataset

- Train model on the source domain **A** (**s = 0**), we view the full source data as a test set.
```python
cd object/
python SP-ASFDA_source.py --trte full --da uda --output ckps/source/ --gpu_id 0 --dset office --max_epoch 100 --s 0 --t 1
```

- Adaptation to other target domains **D and W**, respectively
```python
python SP-ASFDA_target.py --threshold 10 --confidence_threshold 0.5 --da uda --output_src ckps/source/ --output ckps/target/ --gpu_id 0 --dset office --s 0 --t 1  
```

Active Source free domain adaptation (ASFDA) on the VisDA-C dataset

- Synthetic-to-real 
```python
cd object/
 python SP-ASFDA_source.py --trte full --output ckps/source/ --da uda --gpu_id 0 --dset VISDA-C --net resnet101 --lr 1e-3 --max_epoch 10 --s 0 --t 1
 python SP-ASFDA_target.py --threshold 10 --confidence_threshold 0.5 --da uda --dset VISDA-C --gpu_id 0 --s 0 --t 1 --output_src ckps/source/ --output ckps/target/ --net resnet101 --lr 1e-3
 ```