Segmentation sémantique

Segmentation sémantique d’images

La segmentation sémantique associe chaque pixel d’une image à un étiquette de classe telle que fleur, personne, route, ciel ou voiture. Utilisez les applications Image Labeler et Video Labeler pour labéliser les pixels de manière interactive et exporter les données labélisées afin d’entraîner un réseau de neurones.

Input image of a seascape, then a series of cubes representing a deep learning network and a semantic segmented output image of the input.

Applications

Image Labeler	Label images for computer vision applications
Video Labeler	Label video for computer vision applications

Fonctions

développer tout

Charger les données d’apprentissage

`combine`	Combine data from multiple datastores
`countEachLabel`	Count occurrence of pixel or box labels
`groundTruth`	Ground truth label data
`imageDatastore`	Datastore for image data
`pixelLabelDatastore`	Datastore for pixel label data
`pixelLabelTrainingData`	Create training data for semantic segmentation from ground truth

Augmenter et prétraiter les données d’apprentissage

`balancePixelLabels`	Balance pixel labels by oversampling block locations in large images (depuis R2020a)
`imwarp`	Apply geometric transformation to image
`imcrop`	Crop image
`imresize`	Resize image
`transform`	Transform datastore
`randomAffine2d`	Create randomized 2-D affine transformation (depuis R2019b)
`randomWindow2d`	Randomly select rectangular region in image (depuis R2021a)
`centerCropWindow2d`	Create rectangular center cropping window (depuis R2019b)

Concevoir des réseaux de Deep Learning pour la segmentation sémantique

`unet`	Create U-Net convolutional neural network for semantic segmentation (depuis R2024a)
`unet3d`	Create 3-D U-Net convolutional neural network for semantic segmentation of volumetric images (depuis R2024a)
`deeplabv3plus`	Create DeepLab v3+ convolutional neural network for semantic image segmentation (depuis R2024a)
`focalCrossEntropy`	Compute focal cross-entropy loss (depuis R2020b)

Segmenter les images avec le Deep Learning

`semanticseg`	Semantic image segmentation using deep learning
`segmentAnythingModel`	Pretrained Segment Anything Model (SAM) for semantic segmentation (depuis R2024a)

Visualiser la segmentation

`labeloverlay`	Overlay label matrix regions on 2-D image
`volshow`	Display volume
`insertObjectMask`	Insert masks in image or video stream (depuis R2020b)

Évaluer les résultats de la segmentation

`evaluateSemanticSegmentation`	Evaluate semantic segmentation data set against ground truth
`bfscore`	Contour matching score for image segmentation
`dice`	Sørensen-Dice similarity coefficient for image segmentation
`generalizedDice`	Generalized Sørensen-Dice similarity coefficient for image segmentation (depuis R2021a)
`jaccard`	Jaccard similarity coefficient for image segmentation
`segmentationConfusionMatrix`	Confusion matrix of multi-class pixel-level image segmentation (depuis R2020b)
`semanticSegmentationMetrics`	Semantic segmentation quality metrics

Rubriques

En savoir plus

Label Pixels for Semantic Segmentation
Label pixels for training a semantic segmentation network by using a labeling app.
How Labeler Apps Store Exported Pixel Labels
Learn how the labeling apps store pixel label data.
Choose Function to Visualize Detected Objects
Compare visualization functions.
Getting Started with Semantic Segmentation Using Deep Learning
Segment objects by class using deep learning.
Getting Started with Point Clouds Using Deep Learning
Understand how to use point clouds for deep learning.

Créer des données d’apprentissage pour la segmentation sémantique

Datastores for Deep Learning (Deep Learning Toolbox)
Learn how to use datastores in deep learning applications.
Training Data for Object Detection and Semantic Segmentation
Create training data for object detection or semantic segmentation using the Image Labeler or Video Labeler.
Get Started with Image Preprocessing and Augmentation for Deep Learning
Preprocess data for deep learning applications with deterministic operations such as resizing, or augment training data with randomized operations such as random cropping.