FIDs

FIDs is the plural form of the acronym FID, used in several technical domains to denote different concepts. The most widely encountered meanings today are in machine learning and in geographic information systems, though other fields may use the same letters to refer to separate identifiers.

Frechet Inception Distance (FID) is a metric for evaluating the realism of images produced by generative models such as GANs. It measures how close the distribution of features from generated images is to the distribution from real images, computing the Fréchet distance between two multivariate Gaussians fitted to features extracted by a pretrained Inception network. Lower FID indicates closer similarity to real data. The measure was introduced by Martin Heusel, Hubert Ramsauer, Stefan Unterthiner, Johannes Nessler, and Sepp Hochreiter in 2017. In practice, FID requires careful preprocessing and a fixed feature space, and while it often correlates with human judgments, it has limitations: it can be sensitive to sample size, dataset biases, and the choice of network used for feature extraction.

Feature ID (FID) is the internal unique identifier assigned to each feature in a geographic dataset. In formats such as ESRI shapefiles and related GIS databases, the FID provides a stable reference to a feature and is automatically generated when features are created. It is typically separate from user-defined attribute fields and may not be preserved when data are exported to other formats. While useful for software reference and editing operations, the FID is generally not intended as a user-editable data attribute and may vary across transformations or revisions of a dataset.

Other uses of FIDs exist in various domains, so the exact meaning should be inferred from context.