Module: tfm.vision.configs.semantic_segmentation

Semantic segmentation configuration definition.

Classes

class DataConfig: Input config for training.

class DenseFeatureConfig: Config for dense features, such as RGB pixels, masks, heatmaps.

class Evaluation: Evaluation config.

class ExportConfig: Model export config.

class Losses: Loss function config.

class MaskScoringHead: Mask Scoring head config.

class SegmentationHead: Segmentation head config.

class SemanticSegmentationModel: Semantic segmentation model config.

class SemanticSegmentationTask: The model config.

Functions

mnv2_deeplabv3_cityscapes(...): Image segmentation on cityscapes with mobilenetv2 deeplabv3.

mnv2_deeplabv3_pascal(...): Image segmentation on pascal with mobilenetv2 deeplabv3.

mnv2_deeplabv3plus_cityscapes(...): Image segmentation on cityscapes with mobilenetv2 deeplabv3plus.

seg_deeplabv3_pascal(...): Image segmentation on pascal voc with resnet deeplabv3.

seg_deeplabv3plus_cityscapes(...): Image segmentation on cityscapes with resnet deeplabv3+.

seg_deeplabv3plus_pascal(...): Image segmentation on pascal voc with resnet deeplabv3+.

seg_resnetfpn_pascal(...): Image segmentation on pascal voc with resnet-fpn.

semantic_segmentation(...): Semantic segmentation general.

CITYSCAPES_INPUT_PATH_BASE 'cityscapes'
CITYSCAPES_TRAIN_EXAMPLES 2975
CITYSCAPES_VAL_EXAMPLES 500
PASCAL_INPUT_PATH_BASE 'gs://**/pascal_voc_seg'
PASCAL_TRAIN_EXAMPLES 10582
PASCAL_VAL_EXAMPLES 1449