Struct DedupConfig 

pub struct DedupConfig {

    pub database_path: PathBuf,
    pub perceptual_threshold: f64,
    pub ssim_threshold: f64,
    pub histogram_threshold: f64,
    pub feature_match_threshold: usize,
    pub audio_threshold: f64,
    pub metadata_threshold: f64,
    pub parallel: bool,
    pub sample_frames: usize,
    pub chunk_size: usize,
    pub thumbnail_resolution: usize,
    pub bloom_prescreen: bool,
    pub bloom_capacity: usize,
    pub bloom_fpr: f32,
    pub use_lsh: bool,
    pub lsh_num_tables: usize,
    pub lsh_bits_per_table: usize,
    pub lsh_seed: u64,
}

Configuration for deduplication.

Fields

database_path: PathBuf

Database path

perceptual_threshold: f64

Perceptual hash similarity threshold (0.0-1.0)

ssim_threshold: f64

SSIM similarity threshold (0.0-1.0)

histogram_threshold: f64

Histogram similarity threshold (0.0-1.0)

feature_match_threshold: usize

Feature match threshold (minimum number of matches)

audio_threshold: f64

Audio fingerprint similarity threshold (0.0-1.0)

metadata_threshold: f64

Metadata similarity threshold (0.0-1.0)

parallel: bool

Enable parallel processing

sample_frames: usize

Number of frames to sample for video analysis

chunk_size: usize

Chunk size for content-based chunking (bytes)

thumbnail_resolution: usize

Thumbnail resolution for SSIM duplicate detection.

Specifies both width and height of the grayscale thumbnail used for SSIM comparison. Must be >= 4. Default is 8 (i.e. 8x8 = 64 pixels). Higher values give more accurate SSIM at the cost of storage and CPU.

bloom_prescreen: bool

Enable bloom filter pre-screening before expensive perceptual comparisons.

When enabled, a bloom filter is used to quickly reject items whose content hash is already known to be unique, avoiding expensive pairwise perceptual hash comparisons.

bloom_capacity: usize

Expected capacity for the bloom filter pre-screener.

bloom_fpr: f32

False positive rate for the bloom filter pre-screener.

use_lsh: bool

Use LSH acceleration for perceptual hash deduplication.

When enabled, find_perceptual_duplicates() uses a BitLshIndex instead of O(n^2) pairwise comparison. This provides sub-quadratic performance for large libraries at the cost of slightly reduced recall.

lsh_num_tables: usize

Number of LSH hash tables (more = better recall, more memory).

lsh_bits_per_table: usize

Bits sampled per LSH table (fewer = more candidates = better recall).

lsh_seed: u64

Deterministic seed for LSH projections.

Trait Implementations

impl Clone for DedupConfig

fn clone(&self) -> DedupConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for DedupConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for DedupConfig

fn default() -> Self

Returns the “default value” for a type.