ObjectDetectionOptions<T>

Options & Configuration · AiDotNet.Models.Options

Configuration options for object detection models.

Inheritance: ModelOptions → ObjectDetectionOptions<T>

Known derived types: DINOOptions<T>, RTDETROptions<T>

For Beginners

Object detection finds and locates objects in images. This class configures how the detection model works, including:

  • Which architecture to use (YOLO, DETR, Faster R-CNN, etc.)
  • Model size (smaller = faster, larger = more accurate)
  • Detection thresholds (how confident the model must be)

Constructors

Constructor Summary
ObjectDetectionOptions() Creates options with the historical default seed of 42.

Properties

Property Summary
BackboneType Backbone { get; set; } The backbone network type for feature extraction.
DetectionArchitecture Architecture { get; set; } The detection architecture to use.
DetrSetLossOptions SetPredictionLoss { get; set; } Set prediction loss used by TrainDetections on DETR-family detectors, or null for the detector’s published recipe (DETR, DINO or RT-DETR).
ModelSize Size { get; set; } The model size variant.
NeckType Neck { get; set; } The neck architecture for multi-scale feature fusion.
TaskAlignedLossOptions TaskAlignedLoss { get; set; } Task-aligned assignment and loss used by TrainDetections on anchor-free YOLO detectors (YOLOv8, YOLOv9, YOLOv10, YOLOv11), or null for the published defaults.
TwoStageDetectionLossOptions TwoStageLoss { get; set; } Proposal and region-of-interest sampling and loss weights used by TrainDetections on Faster R-CNN and Cascade R-CNN, or null for the published defaults.
bool FreezeBackbone { get; set; } Whether to freeze the backbone during fine-tuning.
bool UseMultiScale { get; set; } Whether to use multi-scale inference for better accuracy.
bool UsePretrained { get; set; } Whether to use pre-trained weights (COCO dataset).
double ConfidenceThreshold { get; set; } Minimum confidence score for a detection to be kept.
double NmsThreshold { get; set; } IoU threshold for Non-Maximum Suppression (NMS).
int MaxDetections { get; set; } Maximum number of detections to return per image.
int NumClasses { get; set; } Number of object classes to detect.
int[] InputSize { get; set; } Input image size [height, width] the model expects.
string WeightsUrl { get; set; } Custom URL to download weights from.
string[] ClassNames { get; set; } Class names for the detection classes.