OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Evaluation & Observability

Clean Fid

Clean Fid, distributed as the Python library `clean-fid`, calculates Fréchet Inception Distance (FID) to evaluate generative image models. It supports image-folder comparisons, comparisons against precomputed dataset statistics, and calculations using a generator function. The library also provides CLIP-based FID and Kernel Inception Distance (KID).

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Clean Fid — OpenViST project profile in AI Evaluation & Observability.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when an image-generation evaluation needs FID or related KID scoring and the image preprocessing choices should be handled consistently. It is also useful when comparing image folders or using the documented dataset-statistics and generator workflows.

Who it is for

  • Researchers evaluating generative image models
  • Developers comparing image-generation experiments or model checkpoints
  • Teams seeking consistent image resizing and quantization in FID workflows

Documented uses

  • Calculate FID between two folders of images.
  • Compare generated images with precomputed statistics for a supported dataset.
  • Calculate FID from a generator function and dataset statistics.
  • Calculate CLIP-FID or KID using the library's documented interfaces.
Before you choose

Requirements & limits

The project is open source under the MIT license. Its documented installation uses pip.

  • Different resizing implementations can materially affect FID, so preprocessing consistency is important.
  • JPEG compression can also change FID scores; image encoding should be considered when comparing results.
  • CLIP-FID is selected with the documented CLIP model name option.
  • Precomputed KID statistics are documented for AFHQ, BreCaHAD and MetFaces.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Calculate FID between two image folders

Install clean-fid and calculate FID for two local image folders.

Before you start

  • Python environment with pip
  • Two image folders to compare

Steps

  1. Install the package with the command below.

    Documented command
    pip install clean-fid

  2. Import fid from cleanfid.

  3. Call fid.compute_fid(fdir1, fdir2) with the two folder paths.

Expected outcome

The call returns an FID score for the two image folders.

Guide limitations

  • The recipe is limited to a folder-to-folder FID comparison; it does not specify image preparation or interpretation of the resulting score.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
GaParmar/clean-fid · Official homepage title · retained in its source language
How we review sourcesSuggest a correction