MobileSAM<T>

Models & Types · AiDotNet.ComputerVision.Segmentation.Efficient

MobileSAM: Faster Segment Anything with TinyViT encoder.

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

For Beginners

Mobile segment anything. On-device interactive segmentation.

Common use cases:

  • Mobile segment anything
  • On-device interactive segmentation
  • Lightweight promptable segmentation
  • Mobile AR applications

How It Works

Technical Details:

  • TinyViT encoder distilled from SAM ViT-H
  • Decoupled distillation: image encoder only
  • Same mask decoder as original SAM
  • 60x smaller encoder than SAM-ViT-H

Reference: Zhang et al., “Faster Segment Anything: Towards Lightweight SAM for Mobile Applications”, arXiv 2023.

Example

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

Constructors

Constructor Summary
MobileSAM(NeuralNetworkArchitecture<T> architecture, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction, MobileSAMOptions options) Initializes MobileSAM in native (trainable) mode.
MobileSAM(NeuralNetworkArchitecture<T> architecture, string onnxModelPath, MobileSAMOptions options) Initializes MobileSAM 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)