dataset.py

dataset_params = dict(
    data_path = "/app/data/",
    version = "triplane",
    train_data_loader = dict(
        pickled = True, # If True, first run 'python utils/pickles_generator.py --dataset-config <path to this file> --py-config <path to config.py>' to preprocess the dataset and save as pickles for faster data loading
        phase = "train",
        batch_size = 1,
        shuffle = True,
        num_workers = 12,
        town = ["Town01", "Town03", "Town04", "Town05", "Town06", "Town07", "Town10HD"],
        weather  = ["ClearNoon"],
        vehicle = ["vehicle.tesla.invisible"],
        spawn_point = ["all"],
        step = ["all"],
        selection = ["input_images", "sphere_dataset"],
        factor = 0.08,
        whole_image = True,
        num_imgs = 3,
        depth=True
    ), 
    val_data_loader = dict(
        pickled = False, # True is unsupported
        phase = "test", # "train", "test", "full"
        batch_size = 1,
        shuffle = False,
        num_workers = 12,
        town = ["Town02"],
        weather  = ["ClearNoon"],
        vehicle = ["vehicle.tesla.invisible"],
        spawn_point = [3, 7, 12, 48, 98, 66],
        step = ["all"],
        selection = ["input_images", "sphere_dataset"],
        factor = 0.25,
        depth = True,
    ),    
)