/adapter

Primary LanguageJupyter NotebookApache License 2.0Apache-2.0

1. prepare dataset

prepare shift dataset as follows:

SHIFT2D/
	checkpoints/
    data/   
        pretrain_models/   
        shift/
            continuous/
            distrete/
    mmdet/
...

2. infer on continuous

单卡

python test.py my_shift_cfg/shift_dino_swins_1x.py checkpoints/iter_105792.pth

多卡

bash dist_test.sh my_shift_cfg/shift_dino_swins_1x.py checkpoints/iter_105792.pth

2. adapt on continuous

tent 单卡

python tools/tta_test.py my_shift_cfg/tent/shift_dino_swins_1x_continous_tent.py

tent 多卡

bash tools/tta_dist_test.py my_shift_cfg/tent/shift_dino_swins_1x_continous_tent.py 2

mean teacher 单卡

python tools/tta_test.py my_shift_cfg/meanteacher/shift_dino_swins_1x_continous_tent.py

mean teacher 多卡

bash tools/tta_dist_test.py my_shift_cfg/meanteacher/shift_dino_swins_1x_continous_tent.py 2