open3d.ml.tf.datasets.augment.SemsegAugmentation

class open3d.ml.tf.datasets.augment.SemsegAugmentation(cfg, seed=None)

Class consisting of different augmentation methods for Semantic Segmentation.

Parameters

cfg – Config for augmentation.

ChromaticAutoContrast(feats, cfg)

Improve contrast for RGB features.

Parameters

  • feats – RGB features, should be in range [0-255].

  • cfg – configuration dict.

ChromaticJitter(feats, cfg)

Adds a small noise jitter to features.

Parameters

  • feats – Features.

  • cfg – configuration dict.

ChromaticTranslation(feats, cfg)

Adds a small translation vector to features.

Parameters

  • feats – Features.

  • cfg – configuration dict.

static HueSaturationTranslation(feat, cfg)

Adds small noise to hue and saturation.

Parameters

  • feat – Features.

  • cfg – config dict with keys(‘hue_max’, and ‘saturation_max’).

RandomDropout(pc, feats, labels, cfg)

Randomly drops some points.

Parameters

  • pc – Pointcloud.

  • feats – Features.

  • labels – Labels.

  • cfg – configuration dict.

RandomHorizontalFlip(pc, cfg)

Randomly flips the given axes.

Parameters

  • pc – Pointcloud.

  • cfg – configuraiton dict.

__init__(cfg, seed=None)

Initialize self. See help(type(self)) for accurate signature.

augment(point, feat, labels, cfg, seed=None)