Flyingthings3D Subset

This dataset is part of the Scene Flow datasets provided by the Computer Science Group of the University of Freiburg.

This is a synthetic dataset containing optical flow and disparity, with occlusion, for two cameras (left/right).

• For more information about the dataset, please refer to the CVPR2016 paper.

• To download the dataset, please follow instructions at this link in the section DispNet/FlowNet2.0 dataset subsets.

Flyingthings3DSubsetDataset API

class alodataset.flyingthings3D_subset_dataset.FlyingThings3DSubsetDataset(backward=False, cameras=None, labels=None, **kwargs)

Bases: Generic[torch.utils.data.dataset.T_co]

Synthetic data with flow and disparity

Parameters

backwardbool

if True, read dataset from t=0 to the end, then from the end to t=0.

cameraslist of string

One frame is returned for each camera. Possible cameras : {‘left’, ‘right’}.

labelslist

Labels that will be attached to the frame Possible labels : {‘disp’, ‘disp_occ’, ‘flow’, ‘flow_occ’, ‘flow_backward’, ‘flow_occ_backward’}

**kwargs :

aloscene.BaseDataset parameters

CAMERAS = ['left', 'right']

LABELS = ['disp', 'disp_occ', 'flow', 'flow_occ', 'flow_backward', 'flow_occ_backward']

get_featlabels_element(nb, selected_features, selected_labels, future_past='into_future', include_flow=True)

get_features_labels(selected_features, selected_labels)

get_frames(sequence_data)

getitem(idx)

static id_sf_path(dataset_dir, folder, ext, _id, train_val='train', left_right='left', future_past=None)