Module multi_modal_search

Module multi_modal_search 

Source
Expand description

This module provides a unified interface for multi-modal similarity search, supporting queries across different modalities (text, image, audio, video) with automatic alignment and fusion in a joint embedding space.

§Features

  • Multi-modal queries: Search with text, images, audio, or combinations
  • Cross-modal retrieval: Find images with text queries, or vice versa
  • Hybrid fusion: Combine results from multiple modalities intelligently
  • Production-ready encoders: Real implementations for all modalities
  • SPARQL integration: Query multi-modal RDF data with SPARQL

§Example

use oxirs_vec::multi_modal_search::{MultiModalSearchEngine, MultiModalQuery, QueryModality};

// Create search engine
let engine = MultiModalSearchEngine::new_default()?;

// Text query
let query = MultiModalQuery::text("show me images of cats");
let results = engine.search(&query, 10)?;

// Image query
let image_data = std::fs::read("cat.jpg")?;
let query = MultiModalQuery::image(image_data);
let results = engine.search(&query, 10)?;

// Hybrid query (text + image)
let query = MultiModalQuery::hybrid(vec![
    QueryModality::Text("cute kitten".to_string()),
    QueryModality::Image(image_data),
]);
let results = engine.search(&query, 10)?;

Structs§

MultiModalConfig
Configuration for multi-modal search
MultiModalQuery
A multi-modal query combining one or more modalities
MultiModalSearchEngine
Multi-modal search engine that handles queries across different modalities
MultiModalStatistics
Statistics about the multi-modal search engine
ProductionAudioEncoder
Production audio encoder using MFCC and spectral features
ProductionGraphEncoder
Production graph encoder for knowledge graphs
ProductionImageEncoder
Production image encoder using ResNet-style features
ProductionTextEncoder
Production text encoder using TF-IDF and sentence embeddings
ProductionVideoEncoder
Production video encoder using temporal features
QueryFilter
Filter for query results
SearchResult
Search result from multi-modal query

Enums§

FilterOperator
QueryModality
Query modality with associated data
SearchStrategy
Search strategy for multi-modal queries