Datasets#

Base#

Dataset.

class nerfstudio.data.datasets.base_dataset.InputDataset(dataparser_outputs: DataparserOutputs, scale_factor: float = 1.0, cache_compressed_images: bool = False)[source]#

Dataset that returns images.

Parameters:

  • dataparser_outputs – description of where and how to read input images.

  • scale_factor – The scaling factor for the dataparser outputs

cache_compressed_images#

If cache_compressed_images == True, cache all the image files into RAM in their compressed form (jpeg, png, etc. but not as pytorch tensors)

get_data(image_idx: int, image_type: Literal['uint8', 'float32'] = 'float32') → Dict[source]#

Returns the ImageDataset data as a dictionary.

Parameters:

  • image_idx – The image index in the dataset.

  • image_type – the type of images returned

get_image_float32(image_idx: int) → Float[Tensor, 'image_height image_width num_channels'][source]#

Returns a 3 channel image in float32 torch.Tensor.

Parameters:

image_idx – The image index in the dataset.

get_image_uint8(image_idx: int) → UInt8[Tensor, 'image_height image_width num_channels'][source]#

Returns a 3 channel image in uint8 torch.Tensor.

Parameters:

image_idx – The image index in the dataset.

get_metadata(data: Dict) → Dict[source]#

Method that can be used to process any additional metadata that may be part of the model inputs.

Parameters:

image_idx – The image index in the dataset.

get_numpy_image(image_idx: int) → ndarray[Any, dtype[uint8]][source]#

Returns the image of shape (H, W, 3 or 4).

Parameters:

image_idx – The image index in the dataset.

property image_filenames: List[Path]#

Returns image filenames for this dataset. The order of filenames is the same as in the Cameras object for easy mapping.

SDF Dataset#

SDFStudio dataset.

class nerfstudio.data.datasets.sdf_dataset.SDFDataset(dataparser_outputs: DataparserOutputs, scale_factor: float = 1.0)[source]#

Bases: InputDataset

Dataset that returns images and depths.

Parameters:

  • dataparser_outputs – description of where and how to read input images.

  • scale_factor – The scaling factor for the dataparser outputs.

get_depths_and_normals(depth_filepath: Path, normal_filename: Path, camtoworld: Tensor)[source]#

function to process additional depths and normal information :param depth_filepath: path to depth file :param normal_filename: path to normal file :param camtoworld: camera to world transformation matrix

get_metadata(data: Dict) → Dict[source]#

Method that can be used to process any additional metadata that may be part of the model inputs.

Parameters:

image_idx – The image index in the dataset.

Semantic Dataset#

Semantic dataset.

class nerfstudio.data.datasets.semantic_dataset.SemanticDataset(dataparser_outputs: DataparserOutputs, scale_factor: float = 1.0)[source]#

Bases: InputDataset

Dataset that returns images and semantics and masks.

Parameters:

dataparser_outputs – description of where and how to read input images.

get_metadata(data: Dict) → Dict[source]#

Method that can be used to process any additional metadata that may be part of the model inputs.

Parameters:

image_idx – The image index in the dataset.