YOLOv10<T>
Models & Types · AiDotNet.ComputerVision.Detection.ObjectDetection.YOLO
YOLOv10 object detector with NMS-free training and consistent dual assignments.
Inheritance: ModelBase<T, Tensor<T>, Tensor<T>> → VisionTaskModelBase<T> → ObjectDetectorBase<T> → YOLOv10<T>
Implements: IDetectionTrainingModel<T>
For Beginners
YOLOv10 eliminates the need for Non-Maximum Suppression (NMS) during inference by using consistent dual assignments during training. This makes inference faster and simpler while maintaining accuracy.
How It Works
Key features:
- NMS-free inference via consistent dual assignments
- One-to-one matching during inference (no duplicate predictions)
- One-to-many matching for auxiliary training
- Reduced post-processing latency
Reference: Wang et al., “YOLOv10: Real-Time End-to-End Object Detection”, 2024
Constructors
| Constructor | Summary |
|---|---|
YOLOv10(ObjectDetectionOptions<T> options, bool useNmsFree) |
Creates a new YOLOv10 detector. |
Properties
| Property | Summary |
|---|---|
string Name { get; } |
Methods
| Method | Summary |
|---|---|
DetectionResult<T> Detect(Tensor<T> image, double confidenceThreshold, double nmsThreshold) |
|
Task LoadWeightsAsync(string pathOrUrl, CancellationToken cancellationToken) |
|
double EffectiveNmsThreshold(double requested) |
|
void SaveWeights(string path) |
|
void Train(Tensor<T> input, Tensor<T> expectedOutput) |
Regresses both heads onto a raw output-shaped target and sums the two losses. |
void TrainDetections(Tensor<T> input, DetectionTrainingBatch<T> targets) |
Trains both heads with YOLOv10’s consistent dual assignments. |