AI 용어집
인공지능 완전 사전
Pose Estimation
Computer vision technique that involves predicting the position and orientation of an object or person in space from an image or video.
Keypoint Detection
Process of identifying and locating specific anatomical points (joints, extremities) on a human body or object to facilitate pose analysis.
Human Pose Estimation
Specialized application of pose estimation that focuses on detecting joint positions of the human body to analyze movements and postures.
2D Pose Estimation
Method that determines the coordinates of keypoints in a two-dimensional plane of the image without considering depth or 3D information.
3D Pose Estimation
Advanced technique that reconstructs the complete spatial position of keypoints in three dimensions, including depth and orientation in real space.
Heatmap Regression
Approach where the neural network generates 2D heatmaps indicating the probability of presence of each keypoint at different positions in the image.
Skeleton Tracking
Continuous process of tracking the articulated skeleton across multiple video frames to analyze movements and transitions between poses.
Multi-person Pose Estimation
Technical challenge consisting of simultaneously detecting poses of multiple individuals in the same scene while avoiding confusion between people.
Pose Normalization
Preprocessing step that standardizes detected poses by eliminating variations due to scale, rotation, or translation to facilitate comparative analysis.
Articulation Modeling
Mathematical representation of biomechanical constraints and relationships between joints to ensure anatomically valid poses.
Pose Lifting
Technique that converts a 2D pose estimation into 3D representation using anatomical models and learned depth constraints.
Temporal Pose Estimation
Approach that exploits temporal information from video sequences to improve the accuracy and consistency of pose estimations over time.
Pose-driven Animation
Application where estimated poses serve as control data to generate real-time animations of characters or objects.
Pose-aware Neural Networks
Neural architectures that explicitly integrate information about pose structure and constraints to improve prediction robustness.
Pose Refinement
Iterative process of improving initial pose estimations using kinematic models and physics-based constraint optimizations.
Graphical Models for Pose
Mathematical representation using graphs to model spatial relationships and dependencies between different body parts in pose estimation.
Occlusion Handling
Set of techniques to handle body parts or objects that are hidden in the image, using inferences based on anatomical consistency.
Pose Confidence Score
Quantitative metric indicating the system's level of certainty about the accuracy of each keypoint detected in pose estimation.