├── .gitignore ├── LICENSE ├── README.md ├── configs ├── pretrain_catmae_vit-s-16.json └── pretrain_catmae_vit-s-8.json ├── downstream ├── davis2017-seg │ ├── DAVIS.md │ ├── code │ │ ├── data │ │ │ ├── jhmdb.py │ │ │ ├── kinetics.py │ │ │ ├── video.py │ │ │ └── vos.py │ │ ├── eval │ │ │ ├── convert_davis.py │ │ │ ├── davis_vallist.txt │ │ │ ├── eval_pck.py │ │ │ ├── eval_vip.py │ │ │ ├── palette.py │ │ │ └── run_test.py │ │ ├── model.py │ │ ├── models_mae.py │ │ ├── resnet.py │ │ ├── test.py │ │ ├── test_mae.py │ │ ├── train.py │ │ └── utils │ │ │ ├── __init__.py │ │ │ ├── arguments.py │ │ │ ├── augs.py │ │ │ ├── kornia_augs.py │ │ │ ├── test_utils.py │ │ │ └── visualize.py │ ├── convert_480_880.py │ ├── download_process_data.sh │ └── figs │ │ └── label_propagation.gif └── kinetics-400-action-recognition │ ├── KINETICS400.md │ ├── LICENSE │ ├── datasets.py │ ├── engine_for_finetuning.py │ ├── engine_for_pretraining.py │ ├── environment.yaml │ ├── functional.py │ ├── kinetics.py │ ├── masking_generator.py │ ├── mixup.py │ ├── modeling_finetune.py │ ├── modeling_finetune_2d_vit.py │ ├── modeling_pretrain.py │ ├── optim_factory.py │ ├── rand_augment.py │ ├── random_erasing.py │ ├── run_class_finetuning.py │ ├── run_mae_pretraining.py │ ├── run_videomae_vis.py │ ├── scripts │ └── kinetics │ │ ├── 2d_patch_vit_small_patch16_224 │ │ └── finetune.sh │ │ └── 3d_patch_vit_small_patch16_224 │ │ └── finetune.sh │ ├── ssv2.py │ ├── transforms.py │ ├── utils.py │ ├── video_transforms.py │ ├── vis.sh │ └── volume_transforms.py ├── engine_pretrain.py ├── figures └── arch.png ├── kinetics_dataset.py ├── main_pretrain.py ├── models_catmae.py └── util ├── crop.py ├── datasets.py ├── lars.py ├── lr_decay.py ├── lr_sched.py ├── misc.py └── pos_embed.py /.gitignore: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/.gitignore -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/LICENSE -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/README.md -------------------------------------------------------------------------------- /configs/pretrain_catmae_vit-s-16.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/configs/pretrain_catmae_vit-s-16.json -------------------------------------------------------------------------------- /configs/pretrain_catmae_vit-s-8.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/configs/pretrain_catmae_vit-s-8.json -------------------------------------------------------------------------------- /downstream/davis2017-seg/DAVIS.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/DAVIS.md -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/data/jhmdb.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/data/jhmdb.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/data/kinetics.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/data/kinetics.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/data/video.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/data/video.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/data/vos.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/data/vos.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/eval/convert_davis.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/eval/convert_davis.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/eval/davis_vallist.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/eval/davis_vallist.txt -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/eval/eval_pck.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/eval/eval_pck.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/eval/eval_vip.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/eval/eval_vip.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/eval/palette.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/eval/palette.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/eval/run_test.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/eval/run_test.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/model.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/model.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/models_mae.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/models_mae.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/resnet.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/resnet.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/test.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/test.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/test_mae.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/test_mae.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/train.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/train.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/utils/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/utils/__init__.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/utils/arguments.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/utils/arguments.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/utils/augs.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/utils/augs.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/utils/kornia_augs.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/utils/kornia_augs.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/utils/test_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/utils/test_utils.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/code/utils/visualize.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/code/utils/visualize.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/convert_480_880.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/convert_480_880.py -------------------------------------------------------------------------------- /downstream/davis2017-seg/download_process_data.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/download_process_data.sh -------------------------------------------------------------------------------- /downstream/davis2017-seg/figs/label_propagation.gif: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/davis2017-seg/figs/label_propagation.gif -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/KINETICS400.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/KINETICS400.md -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/LICENSE: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/LICENSE -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/datasets.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/datasets.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/engine_for_finetuning.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/engine_for_finetuning.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/engine_for_pretraining.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/engine_for_pretraining.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/environment.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/environment.yaml -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/functional.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/functional.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/kinetics.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/kinetics.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/masking_generator.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/masking_generator.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/mixup.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/mixup.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/modeling_finetune.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/modeling_finetune.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/modeling_finetune_2d_vit.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/modeling_finetune_2d_vit.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/modeling_pretrain.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/modeling_pretrain.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/optim_factory.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/optim_factory.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/rand_augment.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/rand_augment.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/random_erasing.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/random_erasing.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/run_class_finetuning.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/run_class_finetuning.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/run_mae_pretraining.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/run_mae_pretraining.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/run_videomae_vis.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/run_videomae_vis.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/scripts/kinetics/2d_patch_vit_small_patch16_224/finetune.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/scripts/kinetics/2d_patch_vit_small_patch16_224/finetune.sh -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/scripts/kinetics/3d_patch_vit_small_patch16_224/finetune.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/scripts/kinetics/3d_patch_vit_small_patch16_224/finetune.sh -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/ssv2.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/ssv2.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/transforms.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/transforms.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/utils.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/video_transforms.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/video_transforms.py -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/vis.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/vis.sh -------------------------------------------------------------------------------- /downstream/kinetics-400-action-recognition/volume_transforms.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/downstream/kinetics-400-action-recognition/volume_transforms.py -------------------------------------------------------------------------------- /engine_pretrain.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/engine_pretrain.py -------------------------------------------------------------------------------- /figures/arch.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/figures/arch.png -------------------------------------------------------------------------------- /kinetics_dataset.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/kinetics_dataset.py -------------------------------------------------------------------------------- /main_pretrain.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/main_pretrain.py -------------------------------------------------------------------------------- /models_catmae.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/models_catmae.py -------------------------------------------------------------------------------- /util/crop.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/util/crop.py -------------------------------------------------------------------------------- /util/datasets.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/util/datasets.py -------------------------------------------------------------------------------- /util/lars.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/util/lars.py -------------------------------------------------------------------------------- /util/lr_decay.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/util/lr_decay.py -------------------------------------------------------------------------------- /util/lr_sched.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/util/lr_sched.py -------------------------------------------------------------------------------- /util/misc.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/util/misc.py -------------------------------------------------------------------------------- /util/pos_embed.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minhoooo1/CatMAE/HEAD/util/pos_embed.py --------------------------------------------------------------------------------