Datasets:
Dataset description (and colab) is missing references to the label aspects of the dataset.
The NYU Depth V2 dataset contains per-image labels including semantic labels, instance information via labels, accelerometer data via accelData, and scene labels: https://cs.nyu.edu/~silberman/datasets/nyu_depth_v2.html. These are not currently captured by the description.
Thanks for looking into this @nathansilberman !
We used a slightly preprocessed version of the dataset from https://github.com/dwofk/fast-depth and that is why we didn't capture those fields. Implementation-wise, it seemed a bit easier. Also, as an FYI, TensorFlow Datasets also follow the same version we used: https://www.tensorflow.org/datasets/catalog/nyu_depth_v2
But I can work on it to switch to the original dataset if that fits better.
Just to clarify: The semantic segmentation and instance information labels are not only missing in the dataset card description, but also in the actual data? So the only labels provided is are the depth maps?