├── Data_Split.py ├── LICENSE ├── Process_Dataset ├── generated_2000videos_captions.json └── readme.md ├── README.md ├── datasets └── readme.txt ├── datasets_part └── readme.txt ├── evaluation ├── generated_captions1.json ├── generated_captions2.json ├── ground_truth_captions1.json ├── ground_truth_captions2.json ├── json2coco.py ├── pycocoevaluationmetric.py ├── test.json └── tsv2coco.py ├── images ├── 1_accident_2.png ├── 2_accident_2.png ├── A1.png ├── A2.png ├── AVD2.png ├── AVD2R.png ├── AVD2Results.png ├── Authors.png ├── EMMAU_accident_example.png ├── FrameWork.png ├── Framework.png ├── Understanding_ability.png └── teaser.png ├── output └── checkpoint │ ├── pred.EMMAU.testing_28frames.beam1.max105.lineidx │ ├── pred.EMMAU.testing_28frames.beam1.max105.lineidx.8b │ ├── pred.EMMAU.testing_28frames.beam1.max105.tsv │ ├── pred.EMMAU.testing_28frames.beam1.max105_0_4.lineidx.8b │ ├── pred.EMMAU.testing_28frames.beam1.max105_1_4.lineidx.8b │ ├── pred.EMMAU.testing_28frames.beam1.max105_2_4.lineidx.8b │ ├── pred.EMMAU.testing_28frames.beam1.max105_3_4.lineidx.8b │ ├── pred.EMMAU_des.testing_28frames.beam1.max105.eval.json │ ├── pred.EMMAU_des.testing_28frames.beam1.max105_coco_format.json │ ├── pred.EMMAU_exp.testing_28frames.beam1.max105.eval.json │ └── pred.EMMAU_exp.testing_28frames.beam1.max105_coco_format.json ├── requirement.txt ├── scripts ├── BDDX_multitask.sh ├── BDDX_multitask_signal_test.sh ├── BDDX_only_caption.sh ├── BDDX_only_signal.sh ├── BDDX_test.sh ├── BDDX_visualize_attn.sh ├── inference.sh └── other_scripts │ ├── BDDX_des_test.sh │ ├── BDDX_des_train.sh │ ├── BDDX_exp_train.sh │ ├── BDDX_multitask_des.sh │ ├── BDDX_multitask_exp.sh │ ├── BDDX_two_train_add_tensor.sh │ ├── BDDX_two_train_swap.sh │ └── BDDX_two_train_wo_cross.sh └── src ├── Super_resolution ├── Super_Resolution.py ├── realesrgan_utils.py └── requirement.txt.txt ├── configs ├── VidSwinBert │ ├── BDDX_8frm_default.json │ ├── BDDX_multi_default.json │ └── BDDX_two_default.json ├── __pycache__ │ └── config.cpython-38.pyc └── config.py ├── datasets ├── __pycache__ │ ├── caption_tensorizer.cpython-38.pyc │ ├── data_sampler.cpython-38.pyc │ ├── sampler_utils.cpython-38.pyc │ ├── vision_language_tsv.cpython-38.pyc │ └── vl_dataloader.cpython-38.pyc ├── caption_tensorizer.py ├── data_sampler.py ├── data_utils │ ├── __pycache__ │ │ ├── image_ops.cpython-38.pyc │ │ ├── video_decoder.cpython-38.pyc │ │ ├── video_functional.cpython-38.pyc │ │ ├── video_ops.cpython-38.pyc │ │ ├── video_transforms.cpython-38.pyc │ │ └── volume_transforms.cpython-38.pyc │ ├── image_ops.py │ ├── video_decoder.py │ ├── video_functional.py │ ├── video_ops.py │ ├── video_transforms.py │ └── volume_transforms.py ├── sampler_utils.py ├── vision_language_tsv.py └── vl_dataloader.py ├── evalcap ├── Readme.md ├── caption_eval.py ├── cider │ ├── cidereval.ipynb │ ├── cidereval.py │ └── data │ │ ├── abstract48S.json │ │ ├── abstract_candsB.json │ │ └── abstract_candsC.json └── utils_caption_evaluate.py ├── modeling ├── __pycache__ │ ├── llama.cpython-38.pyc │ ├── llama_adapter.cpython-38.pyc │ ├── load_bert.cpython-38.pyc │ ├── load_sensor_pred_head.cpython-38.pyc │ ├── load_swin.cpython-38.pyc │ ├── multitask_e2e_vid_swin_bert.cpython-38.pyc │ ├── tokenizer.cpython-38.pyc │ ├── utils.cpython-38.pyc │ └── video_captioning_e2e_vid_swin_bert.cpython-38.pyc ├── llama.py ├── llama_adapter.py ├── load_bert.py ├── load_sensor_pred_head.py ├── load_swin.py ├── multitask_e2e_vid_swin_bert.py ├── signal_predicting.py ├── swin │ ├── __init__.py │ ├── build.py │ ├── config.py │ ├── swin_base_patch4_window12_384.yaml │ ├── swin_base_patch4_window7_224.yaml │ ├── swin_base_patch4_window7_224_22k.yaml │ ├── swin_large_patch4_window12_384.yaml │ ├── swin_large_patch4_window7_224.yaml │ ├── swin_small_patch4_window7_224.yaml │ ├── swin_tiny_patch4_window7_224.yaml │ └── swin_transformer.py ├── tokenizer.py ├── utils.py ├── video_captioning_e2e_vid_swin_bert.py └── video_swin │ ├── __pycache__ │ ├── config.cpython-38.pyc │ └── swin_transformer.cpython-38.pyc │ ├── config.py │ ├── default_runtime.py │ ├── swin_base.py │ ├── swin_base_patch244_window1677_sthv2.py │ ├── swin_base_patch244_window877_kinetics400_1k.py │ ├── swin_base_patch244_window877_kinetics400_22k.py │ ├── swin_base_patch244_window877_kinetics600_22k.py │ ├── swin_large.py │ ├── swin_large_384_patch244_window81212_kinetics400_22k.py │ ├── swin_large_384_patch244_window81212_kinetics600_22k.py │ ├── swin_large_patch244_window877_kinetics400_22k.py │ ├── swin_small_patch244_window877_kinetics400_1k.py │ ├── swin_tiny.py │ ├── swin_tiny_patch244_window877_kinetics400_1k.py │ └── swin_transformer.py ├── prepro ├── LICENSE ├── create_image_frame_tsv.py ├── extract_frames.py ├── preprocess.sh └── tsv_preproc_BDDX.py └── readme.md /Data_Split.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/Data_Split.py -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/LICENSE -------------------------------------------------------------------------------- /Process_Dataset/generated_2000videos_captions.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/Process_Dataset/generated_2000videos_captions.json -------------------------------------------------------------------------------- /Process_Dataset/readme.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/Process_Dataset/readme.md -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/README.md -------------------------------------------------------------------------------- /datasets/readme.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/datasets/readme.txt -------------------------------------------------------------------------------- /datasets_part/readme.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/datasets_part/readme.txt -------------------------------------------------------------------------------- /evaluation/generated_captions1.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/evaluation/generated_captions1.json -------------------------------------------------------------------------------- /evaluation/generated_captions2.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/evaluation/generated_captions2.json -------------------------------------------------------------------------------- /evaluation/ground_truth_captions1.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/evaluation/ground_truth_captions1.json -------------------------------------------------------------------------------- /evaluation/ground_truth_captions2.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/evaluation/ground_truth_captions2.json -------------------------------------------------------------------------------- /evaluation/json2coco.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/evaluation/json2coco.py -------------------------------------------------------------------------------- /evaluation/pycocoevaluationmetric.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/evaluation/pycocoevaluationmetric.py -------------------------------------------------------------------------------- /evaluation/test.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/evaluation/test.json -------------------------------------------------------------------------------- /evaluation/tsv2coco.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/evaluation/tsv2coco.py -------------------------------------------------------------------------------- /images/1_accident_2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/1_accident_2.png -------------------------------------------------------------------------------- /images/2_accident_2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/2_accident_2.png -------------------------------------------------------------------------------- /images/A1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/A1.png -------------------------------------------------------------------------------- /images/A2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/A2.png -------------------------------------------------------------------------------- /images/AVD2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/AVD2.png -------------------------------------------------------------------------------- /images/AVD2R.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/AVD2R.png -------------------------------------------------------------------------------- /images/AVD2Results.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/AVD2Results.png -------------------------------------------------------------------------------- /images/Authors.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/Authors.png -------------------------------------------------------------------------------- /images/EMMAU_accident_example.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/EMMAU_accident_example.png -------------------------------------------------------------------------------- /images/FrameWork.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/FrameWork.png -------------------------------------------------------------------------------- /images/Framework.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/Framework.png -------------------------------------------------------------------------------- /images/Understanding_ability.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/Understanding_ability.png -------------------------------------------------------------------------------- /images/teaser.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/images/teaser.png -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105.lineidx: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105.lineidx -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105.lineidx.8b: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105.lineidx.8b -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105.tsv: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105.tsv -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105_0_4.lineidx.8b: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105_0_4.lineidx.8b -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105_1_4.lineidx.8b: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105_1_4.lineidx.8b -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105_2_4.lineidx.8b: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105_2_4.lineidx.8b -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105_3_4.lineidx.8b: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU.testing_28frames.beam1.max105_3_4.lineidx.8b -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU_des.testing_28frames.beam1.max105.eval.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU_des.testing_28frames.beam1.max105.eval.json -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU_des.testing_28frames.beam1.max105_coco_format.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU_des.testing_28frames.beam1.max105_coco_format.json -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU_exp.testing_28frames.beam1.max105.eval.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU_exp.testing_28frames.beam1.max105.eval.json -------------------------------------------------------------------------------- /output/checkpoint/pred.EMMAU_exp.testing_28frames.beam1.max105_coco_format.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/output/checkpoint/pred.EMMAU_exp.testing_28frames.beam1.max105_coco_format.json -------------------------------------------------------------------------------- /requirement.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/requirement.txt -------------------------------------------------------------------------------- /scripts/BDDX_multitask.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/BDDX_multitask.sh -------------------------------------------------------------------------------- /scripts/BDDX_multitask_signal_test.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/BDDX_multitask_signal_test.sh -------------------------------------------------------------------------------- /scripts/BDDX_only_caption.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/BDDX_only_caption.sh -------------------------------------------------------------------------------- /scripts/BDDX_only_signal.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/BDDX_only_signal.sh -------------------------------------------------------------------------------- /scripts/BDDX_test.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/BDDX_test.sh -------------------------------------------------------------------------------- /scripts/BDDX_visualize_attn.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/BDDX_visualize_attn.sh -------------------------------------------------------------------------------- /scripts/inference.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/inference.sh -------------------------------------------------------------------------------- /scripts/other_scripts/BDDX_des_test.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/other_scripts/BDDX_des_test.sh -------------------------------------------------------------------------------- /scripts/other_scripts/BDDX_des_train.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/other_scripts/BDDX_des_train.sh -------------------------------------------------------------------------------- /scripts/other_scripts/BDDX_exp_train.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/other_scripts/BDDX_exp_train.sh -------------------------------------------------------------------------------- /scripts/other_scripts/BDDX_multitask_des.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/other_scripts/BDDX_multitask_des.sh -------------------------------------------------------------------------------- /scripts/other_scripts/BDDX_multitask_exp.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/other_scripts/BDDX_multitask_exp.sh -------------------------------------------------------------------------------- /scripts/other_scripts/BDDX_two_train_add_tensor.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/other_scripts/BDDX_two_train_add_tensor.sh -------------------------------------------------------------------------------- /scripts/other_scripts/BDDX_two_train_swap.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/other_scripts/BDDX_two_train_swap.sh -------------------------------------------------------------------------------- /scripts/other_scripts/BDDX_two_train_wo_cross.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/scripts/other_scripts/BDDX_two_train_wo_cross.sh -------------------------------------------------------------------------------- /src/Super_resolution/Super_Resolution.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/Super_resolution/Super_Resolution.py -------------------------------------------------------------------------------- /src/Super_resolution/realesrgan_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/Super_resolution/realesrgan_utils.py -------------------------------------------------------------------------------- /src/Super_resolution/requirement.txt.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/Super_resolution/requirement.txt.txt -------------------------------------------------------------------------------- /src/configs/VidSwinBert/BDDX_8frm_default.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/configs/VidSwinBert/BDDX_8frm_default.json -------------------------------------------------------------------------------- /src/configs/VidSwinBert/BDDX_multi_default.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/configs/VidSwinBert/BDDX_multi_default.json -------------------------------------------------------------------------------- /src/configs/VidSwinBert/BDDX_two_default.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/configs/VidSwinBert/BDDX_two_default.json -------------------------------------------------------------------------------- /src/configs/__pycache__/config.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/configs/__pycache__/config.cpython-38.pyc -------------------------------------------------------------------------------- /src/configs/config.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/configs/config.py -------------------------------------------------------------------------------- /src/datasets/__pycache__/caption_tensorizer.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/__pycache__/caption_tensorizer.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/__pycache__/data_sampler.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/__pycache__/data_sampler.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/__pycache__/sampler_utils.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/__pycache__/sampler_utils.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/__pycache__/vision_language_tsv.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/__pycache__/vision_language_tsv.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/__pycache__/vl_dataloader.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/__pycache__/vl_dataloader.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/caption_tensorizer.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/caption_tensorizer.py -------------------------------------------------------------------------------- /src/datasets/data_sampler.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_sampler.py -------------------------------------------------------------------------------- /src/datasets/data_utils/__pycache__/image_ops.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/__pycache__/image_ops.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/data_utils/__pycache__/video_decoder.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/__pycache__/video_decoder.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/data_utils/__pycache__/video_functional.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/__pycache__/video_functional.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/data_utils/__pycache__/video_ops.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/__pycache__/video_ops.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/data_utils/__pycache__/video_transforms.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/__pycache__/video_transforms.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/data_utils/__pycache__/volume_transforms.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/__pycache__/volume_transforms.cpython-38.pyc -------------------------------------------------------------------------------- /src/datasets/data_utils/image_ops.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/image_ops.py -------------------------------------------------------------------------------- /src/datasets/data_utils/video_decoder.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/video_decoder.py -------------------------------------------------------------------------------- /src/datasets/data_utils/video_functional.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/video_functional.py -------------------------------------------------------------------------------- /src/datasets/data_utils/video_ops.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/video_ops.py -------------------------------------------------------------------------------- /src/datasets/data_utils/video_transforms.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/video_transforms.py -------------------------------------------------------------------------------- /src/datasets/data_utils/volume_transforms.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/data_utils/volume_transforms.py -------------------------------------------------------------------------------- /src/datasets/sampler_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/sampler_utils.py -------------------------------------------------------------------------------- /src/datasets/vision_language_tsv.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/vision_language_tsv.py -------------------------------------------------------------------------------- /src/datasets/vl_dataloader.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/datasets/vl_dataloader.py -------------------------------------------------------------------------------- /src/evalcap/Readme.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/evalcap/Readme.md -------------------------------------------------------------------------------- /src/evalcap/caption_eval.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/evalcap/caption_eval.py -------------------------------------------------------------------------------- /src/evalcap/cider/cidereval.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/evalcap/cider/cidereval.ipynb -------------------------------------------------------------------------------- /src/evalcap/cider/cidereval.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/evalcap/cider/cidereval.py -------------------------------------------------------------------------------- /src/evalcap/cider/data/abstract48S.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/evalcap/cider/data/abstract48S.json -------------------------------------------------------------------------------- /src/evalcap/cider/data/abstract_candsB.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/evalcap/cider/data/abstract_candsB.json -------------------------------------------------------------------------------- /src/evalcap/cider/data/abstract_candsC.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/evalcap/cider/data/abstract_candsC.json -------------------------------------------------------------------------------- /src/evalcap/utils_caption_evaluate.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/evalcap/utils_caption_evaluate.py -------------------------------------------------------------------------------- /src/modeling/__pycache__/llama.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/__pycache__/llama.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/__pycache__/llama_adapter.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/__pycache__/llama_adapter.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/__pycache__/load_bert.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/__pycache__/load_bert.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/__pycache__/load_sensor_pred_head.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/__pycache__/load_sensor_pred_head.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/__pycache__/load_swin.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/__pycache__/load_swin.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/__pycache__/multitask_e2e_vid_swin_bert.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/__pycache__/multitask_e2e_vid_swin_bert.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/__pycache__/tokenizer.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/__pycache__/tokenizer.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/__pycache__/utils.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/__pycache__/utils.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/__pycache__/video_captioning_e2e_vid_swin_bert.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/__pycache__/video_captioning_e2e_vid_swin_bert.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/llama.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/llama.py -------------------------------------------------------------------------------- /src/modeling/llama_adapter.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/llama_adapter.py -------------------------------------------------------------------------------- /src/modeling/load_bert.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/load_bert.py -------------------------------------------------------------------------------- /src/modeling/load_sensor_pred_head.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/load_sensor_pred_head.py -------------------------------------------------------------------------------- /src/modeling/load_swin.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/load_swin.py -------------------------------------------------------------------------------- /src/modeling/multitask_e2e_vid_swin_bert.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/multitask_e2e_vid_swin_bert.py -------------------------------------------------------------------------------- /src/modeling/signal_predicting.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/signal_predicting.py -------------------------------------------------------------------------------- /src/modeling/swin/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/__init__.py -------------------------------------------------------------------------------- /src/modeling/swin/build.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/build.py -------------------------------------------------------------------------------- /src/modeling/swin/config.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/config.py -------------------------------------------------------------------------------- /src/modeling/swin/swin_base_patch4_window12_384.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/swin_base_patch4_window12_384.yaml -------------------------------------------------------------------------------- /src/modeling/swin/swin_base_patch4_window7_224.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/swin_base_patch4_window7_224.yaml -------------------------------------------------------------------------------- /src/modeling/swin/swin_base_patch4_window7_224_22k.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/swin_base_patch4_window7_224_22k.yaml -------------------------------------------------------------------------------- /src/modeling/swin/swin_large_patch4_window12_384.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/swin_large_patch4_window12_384.yaml -------------------------------------------------------------------------------- /src/modeling/swin/swin_large_patch4_window7_224.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/swin_large_patch4_window7_224.yaml -------------------------------------------------------------------------------- /src/modeling/swin/swin_small_patch4_window7_224.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/swin_small_patch4_window7_224.yaml -------------------------------------------------------------------------------- /src/modeling/swin/swin_tiny_patch4_window7_224.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/swin_tiny_patch4_window7_224.yaml -------------------------------------------------------------------------------- /src/modeling/swin/swin_transformer.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/swin/swin_transformer.py -------------------------------------------------------------------------------- /src/modeling/tokenizer.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/tokenizer.py -------------------------------------------------------------------------------- /src/modeling/utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/utils.py -------------------------------------------------------------------------------- /src/modeling/video_captioning_e2e_vid_swin_bert.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_captioning_e2e_vid_swin_bert.py -------------------------------------------------------------------------------- /src/modeling/video_swin/__pycache__/config.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/__pycache__/config.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/video_swin/__pycache__/swin_transformer.cpython-38.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/__pycache__/swin_transformer.cpython-38.pyc -------------------------------------------------------------------------------- /src/modeling/video_swin/config.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/config.py -------------------------------------------------------------------------------- /src/modeling/video_swin/default_runtime.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/default_runtime.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_base.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_base.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_base_patch244_window1677_sthv2.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_base_patch244_window1677_sthv2.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_base_patch244_window877_kinetics400_1k.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_base_patch244_window877_kinetics400_1k.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_base_patch244_window877_kinetics400_22k.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_base_patch244_window877_kinetics400_22k.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_base_patch244_window877_kinetics600_22k.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_base_patch244_window877_kinetics600_22k.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_large.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_large.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_large_384_patch244_window81212_kinetics400_22k.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_large_384_patch244_window81212_kinetics400_22k.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_large_384_patch244_window81212_kinetics600_22k.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_large_384_patch244_window81212_kinetics600_22k.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_large_patch244_window877_kinetics400_22k.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_large_patch244_window877_kinetics400_22k.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_small_patch244_window877_kinetics400_1k.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_small_patch244_window877_kinetics400_1k.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_tiny.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_tiny.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_tiny_patch244_window877_kinetics400_1k.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_tiny_patch244_window877_kinetics400_1k.py -------------------------------------------------------------------------------- /src/modeling/video_swin/swin_transformer.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/modeling/video_swin/swin_transformer.py -------------------------------------------------------------------------------- /src/prepro/LICENSE: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/prepro/LICENSE -------------------------------------------------------------------------------- /src/prepro/create_image_frame_tsv.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/prepro/create_image_frame_tsv.py -------------------------------------------------------------------------------- /src/prepro/extract_frames.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/prepro/extract_frames.py -------------------------------------------------------------------------------- /src/prepro/preprocess.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/prepro/preprocess.sh -------------------------------------------------------------------------------- /src/prepro/tsv_preproc_BDDX.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/prepro/tsv_preproc_BDDX.py -------------------------------------------------------------------------------- /src/readme.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/An-Answer-tree/AVD2/HEAD/src/readme.md --------------------------------------------------------------------------------