PixelLM<T>

Models & Types · AiDotNet.ComputerVision.Segmentation.Referring

PixelLM: Pixel Reasoning with Large Multimodal Model.

Inheritance: NeuralNetworkBase<T> → SegmentationModelBase<T> → ReferringSegmentationBase<T> → PixelLM<T>

For Beginners

Multi-target reasoning segmentation. Complex text-to-mask segmentation.

Common use cases:

  • Multi-target reasoning segmentation
  • Complex text-to-mask segmentation
  • Pixel-level reasoning from language
  • Detailed visual grounding

How It Works

Technical Details:

  • Lightweight pixel decoder with codebook tokens
  • Multi-scale feature integration for segmentation
  • Handles multiple target masks from a single query
  • End-to-end LLM + pixel decoder training

Reference: Ren et al., “PixelLM: Pixel Reasoning with Large Multimodal Model”, CVPR 2024.

Example

// Create a PixelLM model for multi-target reasoning segmentation
// Any CLIP-style ITextEncoder supplies the text embeddings, e.g. a CLIP model from VisionLanguage.Encoders.
ITextEncoder<double> textEncoder = clipModel;
var architecture = new NeuralNetworkArchitecture<double>(
    inputType: InputType.ThreeDimensional,
    taskType: NeuralNetworkTaskType.BinaryClassification,
    inputHeight: 1024, inputWidth: 1024, inputDepth: 3, outputSize: 1);
var model = new PixelLM<double>(architecture, textEncoder);
// Every tunable is now set through the options object; these are the defaults:
var custom = new PixelLM<double>(architecture, textEncoder,
    options: new PixelLMOptions { NumClasses = 1, DropRate = 0 });
            
// Or load a pre-trained ONNX model for pixel-level reasoning
var onnxModel = new PixelLM<double>(architecture, "pixellm.onnx", textEncoder, options: null);

Constructors

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

Methods

Method Summary
List<ReferringSegmentationResult<T>> SegmentVideoFromExpression(Tensor<T> frames, string expression)
ModelMetadata<T> GetModelMetadata() Collects metadata describing this model’s configuration.
ReferringSegmentationResult<T> SegmentFromConversation(Tensor<T> image, IReadOnlyList<ValueTuple<string, string>> conversationHistory, string currentQuery)