Merge branch 'add_feature_vector_dataloader' into simplify_video_doma…

…in_adapter

examples/action_dann_lightn/config.py

@@ -16,7 +16,7 @@
 # Dataset
 # -----------------------------------------------------------------------------
 _C.DATASET = CN()
-_C.DATASET.ROOT = "J:/Datasets/EgoAction/" # "/shared/tale2/Shared"
+_C.DATASET.ROOT = "F:/Datasets/EgoAction/" # "/shared/tale2/Shared"
 _C.DATASET.SOURCE = "EPIC" # dataset options=["EPIC", "GTEA", "ADL", "KITCHEN"]
 _C.DATASET.SRC_TRAINLIST = "epic_D1_train.pkl"
 _C.DATASET.SRC_TESTLIST = "epic_D1_test.pkl"

examples/action_dann_lightn/main.py

@@ -50,7 +50,7 @@ def main():
 
  # ---- setup dataset ----
  seed = cfg.SOLVER.SEED
- source, target, dict_num_classes = VideoDataset.get_source_target(
+ source, target, num_classes = VideoDataset.get_source_target(
  VideoDataset(cfg.DATASET.SOURCE.upper()), VideoDataset(cfg.DATASET.TARGET.upper()), seed, cfg
  )
  dataset = VideoMultiDomainDatasets(
@@ -69,7 +69,7 @@ def main():
  set_seed(seed) # seed_everything in pytorch_lightning did not set torch.backends.cudnn
  print(f"==> Building model for seed {seed} ......")
  # ---- setup model and logger ----
- model, train_params = get_model(cfg, dataset, dict_num_classes)
+ model, train_params = get_model(cfg, dataset, num_classes)
  tb_logger = pl_loggers.TensorBoardLogger(cfg.OUTPUT.TB_DIR, name="seed{}".format(seed))
  checkpoint_callback = ModelCheckpoint(
  # dirpath=full_checkpoint_dir,

examples/action_dann_lightn/model.py

@@ -55,22 +55,22 @@ def get_config(cfg):
 
 
 # Based on https://github.com/criteo-research/pytorch-ada/blob/master/adalib/ada/utils/experimentation.py
-def get_model(cfg, dataset, dict_num_classes):
+def get_model(cfg, dataset, num_classes):
  """
  Builds and returns a model and associated hyper parameters according to the config object passed.
 
  Args:
  cfg: A YACS config object.
  dataset: A multi domain dataset consisting of source and target datasets.
- dict_num_classes (dict): The dictionary of class number for specific dataset.
+ num_classes (dict): The dictionary of class number for specific dataset.
  """
 
  # setup feature extractor
  feature_network, class_feature_dim, domain_feature_dim = get_video_feat_extractor(
- cfg.MODEL.METHOD.upper(), cfg.DATASET.IMAGE_MODALITY, cfg.MODEL.ATTENTION, dict_num_classes
+ cfg.MODEL.METHOD.upper(), cfg.DATASET.IMAGE_MODALITY, cfg.MODEL.ATTENTION, num_classes
  )
  # setup classifier
- classifier_network = ClassNetVideo(input_size=class_feature_dim, dict_n_class=dict_num_classes)
+ classifier_network = ClassNetVideo(input_size=class_feature_dim, dict_n_class=num_classes)
 
  config_params = get_config(cfg)
  train_params = config_params["train_params"]
@@ -100,7 +100,7 @@ def get_model(cfg, dataset, dict_num_classes):
  if cfg.DAN.USERANDOM:
  critic_input_size = cfg.DAN.RANDOM_DIM
  else:
- critic_input_size = domain_feature_dim * dict_num_classes["verb"]
+ critic_input_size = domain_feature_dim * num_classes["verb"]
  critic_network = DomainNetVideo(input_size=critic_input_size)
 
  if cfg.DAN.METHOD == "CDAN":

kale/embed/video_feature_extractor.py

@@ -15,7 +15,7 @@
 from kale.loaddata.video_access import get_image_modality
 
 
-def get_video_feat_extractor(model_name, image_modality, attention, dict_num_classes):
+def get_video_feat_extractor(model_name, image_modality, attention, num_classes):
  """
  Get the feature extractor w/o the pre-trained model and SELayers. The pre-trained models are saved in the path
  ``$XDG_CACHE_HOME/torch/hub/checkpoints/``. For Linux, default path is ``~/.cache/torch/hub/checkpoints/``.
@@ -26,7 +26,7 @@ def get_video_feat_extractor(model_name, image_modality, attention, dict_num_cla
  model_name (string): The name of the feature extractor. (Choices=["I3D", "R3D_18", "R2PLUS1D_18", "MC3_18"])
  image_modality (string): Image type. (Choices=["rgb", "flow", "joint"])
  attention (string): The attention type. (Choices=["SELayerC", "SELayerT", "SELayerCoC", "SELayerMC", "SELayerCT", "SELayerTC", "SELayerMAC"])
- dict_num_classes (dict): The class number of specific dataset. (Default: No use)
+ num_classes (dict): The class number of specific dataset.
 
  Returns:
  feature_network (dictionary): The network to extract features.
@@ -37,7 +37,7 @@ def get_video_feat_extractor(model_name, image_modality, attention, dict_num_cla
 
  rgb, flow = get_image_modality(image_modality)
  # only use verb class when input is image.
- num_classes = dict_num_classes["verb"]
+ num_classes = num_classes["verb"]
 
  attention_list = ["SELayerC", "SELayerT", "SELayerCoC", "SELayerMC", "SELayerCT", "SELayerTC", "SELayerMAC"]
  model_list = ["I3D", "R3D_18", "MC3_18", "R2PLUS1D_18"]

tests/pipeline/test_video_domain_adapter.py

@@ -92,7 +92,7 @@ def test_video_domain_adapter(source_cfg, target_cfg, image_modality, da_method,
  )
 
  # build dataset
- source, target, dict_num_classes = VideoDataset.get_source_target(
+ source, target, num_classes = VideoDataset.get_source_target(
  VideoDataset(cfg.DATASET.SOURCE.upper()), VideoDataset(cfg.DATASET.TARGET.upper()), seed, cfg
  )
 
@@ -119,7 +119,7 @@ def test_video_domain_adapter(source_cfg, target_cfg, image_modality, da_method,
  feature_network = {"rgb": VideoBoringModel(3), "flow": VideoBoringModel(2)}
 
  # setup classifier
- classifier_network = ClassNetVideo(input_size=class_feature_dim, dict_n_class=dict_num_classes)
+ classifier_network = ClassNetVideo(input_size=class_feature_dim, dict_n_class=num_classes)
  train_params = testing_training_cfg["train_params"]
  method_params = {}
  method = domain_adapter.Method(da_method)
@@ -144,7 +144,7 @@ def test_video_domain_adapter(source_cfg, target_cfg, image_modality, da_method,
  if cfg.DAN.USERANDOM:
  critic_input_size = 1024
  else:
- critic_input_size = domain_feature_dim * dict_num_classes["verb"]
+ critic_input_size = domain_feature_dim * num_classes["verb"]
  critic_network = DomainNetVideo(input_size=critic_input_size)
 
  if da_method == "CDAN":