Thuật ngữ AI
Từ điển đầy đủ về Trí tuệ nhân tạo
Scene Inference
Process of deducing semantic and geometric information about an entire scene from partial visual data, relying on contextual models.
Spatial Relationship Modeling
Mathematical representation of geometric constraints and interactions between objects in a scene, such as proximity, orientation, or mutual support.
Scene Graph
Hierarchical data structure where nodes represent objects and edges represent their spatial or semantic relationships, enabling global interpretation of the environment.
Contextual Reasoning
Ability of a system to use prior knowledge about typical scenes to interpret ambiguous or missing elements in an image.
Scene Semantic Segmentation
Pixel-by-pixel classification of an image into coherent semantic categories, integrating understanding of relationships between segmented regions.
Contextual Anomaly Detection
Identification of objects or situations that violate contextual expectations of a given scene, such as a computer in a forest.
Occlusion Prediction
Estimation of invisible parts of objects in an image based on three-dimensional understanding of the scene and relationships between objects.
Contextual Depth Map
Generation of depth information for an entire scene using contextual cues and spatial relationships between objects.
Scene Activity Recognition
Interpretation of ongoing actions and interactions in a scene by collectively analyzing object poses and their relationships.
Generative Scene Model
System capable of creating plausible scene representations by learning statistical distributions of object arrangements and their relationships.
Contextual Modality Fusion
Integration of data from different sources (visual, textual, depth) to enrich the contextual understanding of a scene.
Scene Contrastive Learning
Training technique where the model learns to distinguish plausible object arrangements from implausible ones in a given context.
Spatial Constraint Propagation
Algorithm allowing the propagation of spatial relationship information throughout a scene to refine local predictions.
Holistic Scene Interpretation
Approach considering the scene as an integrated whole rather than a collection of independent objects, emphasizing global emergences.
Affordance Modeling
Detection of interaction possibilities offered by objects in their spatial context, such as the possibility of sitting on a chair.
Scene Graph Neural Network
Deep learning architecture specifically designed to process scene graphs by learning to propagate information through object relationships.
Scene Layout Estimation
Prediction of the overall geometric structure of an indoor environment, including walls, ceilings and floors, using contextual cues.