PIDNet<T>
Models & Types · AiDotNet.ComputerVision.Segmentation.Efficient
PIDNet: A Real-time Semantic Segmentation Network Inspired by PID Controllers.
Inheritance: NeuralNetworkBase<T> → SegmentationModelBase<T> → SemanticSegmentationBase<T> → PIDNet<T>
For Beginners
Real-time semantic segmentation for autonomous driving. Cityscapes and urban scene parsing.
Common use cases:
- Real-time semantic segmentation for autonomous driving
- Cityscapes and urban scene parsing
- Edge-device semantic segmentation
- Balanced speed-accuracy segmentation
How It Works
Technical Details:
- PID-controller-inspired three-branch architecture (P, I, D)
- P-branch: detail features, I-branch: context, D-branch: boundary
- Bag of Tricks for efficient feature fusion
- Achieves Cityscapes mIoU 80+ at 90+ FPS
Reference: Xu et al., “PIDNet: A Real-time Semantic Segmentation Network Inspired by PID Controllers”, CVPR 2023.
Example
var trainingMasks = Tensor<double>.CreateRandom(1, 3, 32, 32);
var inputImage = Tensor<double>.CreateRandom(1, 3, 32, 32);
var trainingImages = Tensor<double>.CreateRandom(4, 3, 32, 32);
// Use AiModelBuilder facade for real-time semantic segmentation
var architecture = new NeuralNetworkArchitecture<double>(
inputType: InputType.ThreeDimensional,
taskType: NeuralNetworkTaskType.MultiClassClassification,
inputHeight: 512, inputWidth: 1024, inputDepth: 3, outputSize: 19);
var builder = new AiModelBuilder<double, Tensor<double>, Tensor<double>>()
.ConfigureModel(new PIDNet<double>(architecture));
var result = builder.Build(trainingImages, trainingMasks);
var segmentation = result.Predict(inputImage);
Constructors
| Constructor | Summary |
|---|---|
PIDNet(NeuralNetworkArchitecture<T> architecture, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction, PIDNetOptions options) |
Initializes PIDNet in native (trainable) mode. |
PIDNet(NeuralNetworkArchitecture<T> architecture, string onnxModelPath, PIDNetOptions options) |
Initializes PIDNet in ONNX (inference-only) mode. |
Methods
| Method | Summary |
|---|---|
ModelMetadata<T> GetModelMetadata() |
Collects metadata describing this model’s configuration. |