AI-ordlista
Den kompletta ordlistan över AI
Object Center
Central point of a detected object, serving as the main reference in CenterNet to infer the complete dimensions of the bounding box.
Size Regression
Prediction branch in CenterNet that estimates the width and height of each object from its detected central point.
Point-based Detection
Detection paradigm that replaces anchor boxes with direct prediction of keypoints, reducing complexity and improving localization accuracy.
Center Offset
Offset vector predicted by CenterNet to adjust the position of an object's center from the discrete grid to its exact continuous location.
Top-left Keypoint
Characteristic point detected in certain CenterNet variants corresponding to the top-left corner of the object's bounding box.
Bottom-right Keypoint
Complementary characteristic point detected to define the bottom-right corner of the bounding box, enabling complete object reconstruction.
Object Keypoint Similarity (OKS)
Evaluation metric that measures the precision of keypoint detection by taking into account the scale and variance of each detected keypoint.
Convolutional Pose Machine
Sequential architecture that influences CenterNet in its ability to progressively refine keypoint predictions through multiple convolutional stages.
Spatial Resolution Preservation
CenterNet's ability to maintain high spatial resolution in feature maps for accurate object center localization.