A table of machine-computable features

Keys built from images (see keyFromTerms() and keyFromClusters()) carry a features table that says how a machine computes the score each lead tests. classify() walks a key over a matrix of such scores.

Usage

featureTable(
  id,
  kind,
  label = id,
  prompt = NA,
  negative_prompt = NA,
  region = NA,
  embedding = NULL,
  exemplars = NULL,
  definition = NA
)

Arguments

id

Feature ids, e.g. "sniglet:dulmzil" or "text:round_headlights".

kind

"centroid_patch_max" (highest cosine similarity between any image patch embedding and embedding), "clip_text_pair" (cosine with prompt minus cosine with negative_prompt) or "external" (computed by the user).

label

Name shown to people; defaults to id.

prompt, negative_prompt

Texts for clip_text_pair features.

region

Crop as "y0,y1,x0,x1" fractions, or NA for the whole image.

embedding

List of numeric vectors (one per feature), or NULL.

exemplars

List of data frames with columns image, patch, score and optionally png (base64) showing what the feature means.

definition

Optional sentence saying what the feature is, shown in the wizard's glossary (see renameSniglets()).

Value

A data frame with the columns above; embedding and exemplars are list columns.

Examples

featureTable("sniglet:x", "centroid_patch_max", embedding = list(c(1, 0, 0)))
#>          id               kind     label prompt negative_prompt region
#> 1 sniglet:x centroid_patch_max sniglet:x   <NA>            <NA>   <NA>
#>   definition embedding exemplars
#> 1       <NA>   1, 0, 0      NULL