search

google-research / kubric

A data generation pipeline for creating semi-realistic synthetic multi-object videos with rich annotations such as instance segmentation masks, depth maps, and optical flow.
Apache License 2.0
2.27k stars 223 forks source link

Some data fields not available in publicly released MOVi dataset #270

Open chaitanya100100 opened 1 year ago

chaitanya100100 commented 1 year ago

I am trying to access publicly released MOVi-B dataset. But some data fields are not available. I did the following:

ds, ds_info = tfds.load("movi_b", data_dir="gs://kubric-public/tfds", with_info=True)
train_iter = iter(tfds.as_numpy(ds["train"]))
example = next(train_iter)
print(example.keys())
print(example['instances'].keys())
print(example['metadata'].keys())

The output is:

dict_keys(['background_color', 'backward_flow', 'camera', 'depth', 'events', 'forward_flow', 'instances', 'metadata', 'normal', 'object_coordinates', 'segmentations', 'video'])

dict_keys(['angular_velocities', 'bbox_frames', 'bboxes', 'bboxes_3d', 'color', 'friction', 'image_positions', 'mass', 'material_label', 'positions', 'quaternions', 'restitution', 'shape_label', 'velocities', 'visibility'])

dict_keys(['backward_flow_range', 'depth_range', 'forward_flow_range', 'height', 'num_frames', 'num_instances', 'video_name', 'width'])

I wanted to access example['instances']['size_label'] but it is not available.

Can you please update the public dataset?

chaitanya100100 commented 1 year ago

An update: it should have 'scale' parameter, not 'size_label' . It's not available in the publically available version.