AI 용어집
인공지능 완전 사전
Fusion of Modalities
Techniques for effectively combining and integrating multiple heterogeneous data sources into a unified representation.
Cross-Modal Learning
Methods that enable learning from one modality to improve performance on a different modality.
Shared Multimodal Representations
Creation of shared representation spaces where different modalities can be compared and manipulated together.
Modal Alignment
Process of semantic matching between elements from different modalities (e.g., words and image regions).
Multimodal Translation
Conversion of data from one modality to another, such as generating text from images or images from text.
Multimodal Attention
Attention mechanisms adapted to dynamically weight and select relevant information across modalities.
Vision and Language
Specialized subfield focusing on the interaction between image and text processing for tasks such as captioning or VQA.
Audio-Visual
Simultaneous and integrated processing of audio and video streams for enhanced contextual understanding.
Self-Supervised Multimodal Learning
Label-free learning techniques leveraging natural correlations between different modalities.
Multimodal Transformers
Transformer-based architectures adapted to simultaneously process multiple types of data.
Multimodal Memory
Memory systems capable of efficiently storing and retrieving complex multimodal information.
Few-Shot Multimodal Learning
Techniques enabling learning with very few examples by leveraging relationships between modalities.