efficientunet_tissue_mask_model¶
Tissue Mask Segmentation Model Architecture.
This module defines a tissue segmentation model based on an EfficientNet-UNet architecture for identifying tissue regions in digital pathology images. The model implements an EfficientNetB0 encoder with a UNet-style decoder and segmentation head for high-resolution tissue segmentation.
Key Components:¶
- Conv2dStaticSamePadding:
Convolutional layer with static same padding.
- MBConvBlock:
Mobile Inverted Residual Bottleneck block.
- EfficientNetEncoder:
EfficientNetB0 encoder for feature extraction.
- Conv2dReLU:
Convolutional block with BatchNorm and ReLU activation.
- UnetDecoderBlock:
Decoder block with skip connections for feature fusion.
- UnetDecoder:
Decoder with skip connections for UNet architecture.
- SegmentationHead:
Final layer for segmentation output.
- EfficientUNetTissueMaskModel:
Main model class implementing encoder-decoder architecture for tissue detection.
Features:¶
ImageNet normalization during preprocessing.
Morphological postprocessing for generating clean tissue masks.
Efficient inference pipeline for batch processing.
Example
>>> from tiatoolbox.models.engine.semantic_segmentor import SemanticSegmentor
>>> segmentor = SemanticSegmentor(model="efficientunet-tissue_mask")
>>> results = segmentor.run(
... ["/example_wsi.svs"],
... masks=None,
... auto_get_mask=False,
... patch_mode=False,
... save_dir=Path("/tissue_mask/"),
... output_type="annotationstore",
... )
Classes
Conv2d + BatchNorm + ReLU block. |
|
2D Convolution with static same padding. |
|
EfficientNetB0 encoder for feature extraction. |
|
EfficientNet-UNet Tissue Segmentation Model. |
|
Mobile Inverted Residual Bottleneck block. |
|
Segmentation head for UNet architecture. |
|
UNet decoder with skip connections. |
|
Decoder block for UNet architecture. |