Get Started
Home
Topics
Search
Library
Research questionHow can image retrieval rankings capture neighborhood context in high-dimensional feature spaces?High-dimensional visual features can make pairwise distances a poor guide to which images are meaningfully similar. Rankings may therefore miss relationships among neighboring images and the connection between visual features and high-level semantics.
Computer Vision
Image & Video Processing
Information Retrieval
Machine Learning
Latest papersRecent research connected to this question, newest first.Aggregating Neighbor Embedding Projection and Rank-Based Manifold Learning for Image RetrievalThe source concerns content-based image retrieval using deep features from ResNet152, Swin Transformer, and DINOv2. Its evidence comes from experiments on several public datasets, where rank aggregation combines UMAP-based projections with rank-based manifold-learning refinements and often improves top-ranked precision and retrieval effectiveness.research paper · Sep 2, 2026
Related questions
How can multimodal retrieval distinguish correct attribute–object bindings when images share the same concepts?When can a random projection preserve JL distance bounds yet lose geometric variation needed for rankings and inference?How can image-text retrieval focus on caption-described attributes while ignoring unmentioned visual information?How should landmark-derived features preserve task-relevant geometry under nuisance variation in structured image classification?