Computes mean embedding using Welford's online algorithm, processing one tile at a time without storing all embeddings in memory. Ideal for large regions with many tiles.
Usage
summary_mean_streaming(
gt,
tiles_df,
year,
region = NULL,
sample_rate = 1,
seed = NULL,
progress = TRUE
)Arguments
- gt
GeoTessera object
- tiles_df
Data frame of tile metadata
- year
Integer year
- region
sf object for masking (optional)
- sample_rate
Fraction of pixels to sample (0-1). Default 1.0 (all pixels). Use lower values for faster processing of very large regions.
- seed
Random seed for sampling (if sample_rate < 1)
- progress
Show progress
Examples
if (FALSE) { # \dontrun{
gt <- geotessera()
tiles_df <- gt$get_tiles(bbox, year = 2024)
result <- summary_mean_streaming(gt, tiles_df, 2024)
} # }