RepViTSAM<T>

Models & Types · AiDotNet.ComputerVision.Segmentation.Efficient

RepViT-SAM: Real-time SAM with RepViT backbone.

Inheritance: NeuralNetworkBase<T> → SegmentationModelBase<T> → PromptableSegmentationBase<T> → RepViTSAM<T>

For Beginners

Real-time segment anything. Mobile and edge SAM inference.

Common use cases:

  • Real-time segment anything
  • Mobile and edge SAM inference
  • Interactive segmentation on devices
  • Low-latency promptable segmentation

How It Works

Technical Details:

  • RepViT (Re-parameterized Vision Transformer) backbone
  • Structural re-parameterization for inference speed
  • Multi-scale feature extraction for SAM decoder
  • Achieves real-time SAM on mobile devices

Reference: Wang et al., “RepViT-SAM: Towards Real-Time Segmenting Anything”, arXiv 2024.

Example

// Create a RepViT-SAM model for real-time promptable segmentation
var architecture = new NeuralNetworkArchitecture<double>(
    inputType: InputType.ThreeDimensional,
    taskType: NeuralNetworkTaskType.BinaryClassification,
    inputHeight: 1024, inputWidth: 1024, inputDepth: 3, outputSize: 1);
var model = new RepViTSAM<double>(architecture);
// Every tunable is now set through the options object; these are the defaults:
var custom = new RepViTSAM<double>(architecture,
    options: new RepViTSAMOptions { NumClasses = 1, DropRate = 0 });

Constructors

Constructor Summary
RepViTSAM(NeuralNetworkArchitecture<T> architecture, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction, RepViTSAMOptions options) Initializes RepViTSAM in native (trainable) mode.
RepViTSAM(NeuralNetworkArchitecture<T> architecture, string onnxModelPath, RepViTSAMOptions options) Initializes RepViTSAM in ONNX (inference-only) mode.

Methods

Method Summary
List<PromptedSegmentationResult<T>> SegmentEverything()
ModelMetadata<T> GetModelMetadata() Collects metadata describing this model’s configuration.
PromptedSegmentationResult<T> SegmentFromBox(Tensor<T> box)
PromptedSegmentationResult<T> SegmentFromMask(Tensor<T> mask)
PromptedSegmentationResult<T> SegmentFromPoints(Tensor<T> points, Tensor<T> labels)
void SetImage(Tensor<T> image)