FID calculation with proper image resizing and quantization steps

Overview

clean-fid: Fixing Inconsistencies in FID


Project | Paper

The FID calculation involves many steps that can produce inconsistencies in the final metric. As shown below, different implementations use different low-level image quantization and resizing functions, the latter of which are often implemented incorrectly.

We provide an easy-to-use library to address the above issues and make the FID scores comparable across different methods, papers, and groups.

FID Steps


On Buggy Resizing Libraries and Surprising Subtleties in FID Calculation
Gaurav Parmar, Richard Zhang, Jun-Yan Zhu
arXiv 2104.11222, 2021
CMU and Adobe



Buggy Resizing Operations

The definitions of resizing functions are mathematical and should never be a function of the library being used. Unfortunately, implementations differ across commonly-used libraries. They are often implemented incorrectly by popular libraries.


The inconsistencies among implementations can have a drastic effect of the evaluations metrics. The table below shows that FFHQ dataset images resized with bicubic implementation from other libraries (OpenCV, PyTorch, TensorFlow, OpenCV) have a large FID score (≥ 6) when compared to the same images resized with the correctly implemented PIL-bicubic filter. Other correctly implemented filters from PIL (Lanczos, bilinear, box) all result in relatively smaller FID score (≤ 0.75).

JPEG Image Compression

Image compression can have a surprisingly large effect on FID. Images are perceptually indistinguishable from each other but have a large FID score. The FID scores under the images are calculated between all FFHQ images saved using the corresponding JPEG format and the PNG format.

Below, we study the effect of JPEG compression for StyleGAN2 models trained on the FFHQ dataset (left) and LSUN outdoor Church dataset (right). Note that LSUN dataset images were collected with JPEG compression (quality 75), whereas FFHQ images were collected as PNG. Interestingly, for LSUN dataset, the best FID score (3.48) is obtained when the generated images are compressed with JPEG quality 87.


Quick Start

  • install requirements

    pip install -r requirements.txt
    
  • install the library

    pip install clean-fid
    
  • Compute FID between two image folders

    from cleanfid import fid
    
    score = fid.compute_fid(fdir1, fdir2)
    
  • Compute FID between one folder of images and pre-computed datasets statistics (e.g., FFHQ)

    from cleanfid import fid
    
    score = fid.compute_fid(fdir1, dataset_name="FFHQ", dataset_res=1024)
    
    
  • Compute FID using a generative model and pre-computed dataset statistics:

    from cleanfid import fid
    
    # function that accepts a latent and returns an image in range[0,255]
    gen = lambda z: GAN(latent=z, ... , <other_flags>)
    
    score = fid.compute_fid(gen=gen, dataset_name="FFHQ",
            dataset_res=256, num_gen=50_000)
    
    

Supported Precomputed Datasets

We provide precompute statistics for the following configurations

Task Dataset Resolution split mode
Image Generation FFHQ 256,1024 train+val clean, legacy_pytorch, legacy_tensorflow
Image Generation LSUN Outdoor Churches 256 train clean, legacy_pytorch, legacy_tensorflow
Image to Image horse2zebra 128,256 train, test, train+test clean, legacy_pytorch, legacy_tensorflow

Using precomputed statistics In order to compute the FID score with the precomputed dataset statistics, use the corresponding options. For instance, to compute the clean-fid score on generated 256x256 FFHQ images use the command:

fid_score = fid.compute_fid(fdir1, dataset_name="FFHQ", dataset_res=256,  mode="clean")

Create Custom Dataset Statistics

  • dataset_path: folder where the dataset images are stored
  • Generate and save the inception statistics
    import numpy as np
    from cleanfid import fid
    dataset_path = ...
    feat = fid.get_folder_features(dataset_path, num=50_000)
    mu = np.mean(feats, axis=0)
    sigma = np.cov(feats, rowvar=False)
    np.savez_compressed("stats.npz", mu=mu, sigma=sigma)
    

Backwards Compatibility

We provide two flags to reproduce the legacy FID score.

  • mode="legacy_pytorch"
    This flag is equivalent to using the popular PyTorch FID implementation provided here
    The difference between using clean-fid with this option and code is ~1.9e-06
    See doc for how the methods are compared

  • mode="legacy_tensorflow"
    This flag is equivalent to using the official implementation of FID released by the authors. To use this flag, you need to additionally install tensorflow. The tensorflow cuda version may cause issues with the pytorch code. I have tested this with TensorFlow-cpu 2.2 (`pip install tensorflow-cpu==2.2)


CleanFID Leaderboard for common tasks


FFHQ @ 1024x1024

Model Legacy-FID Clean-FID
StyleGAN2 2.85 ± 0.05 3.08 ± 0.05
StyleGAN 4.44 ± 0.04 4.82 ± 0.04
MSG-GAN 6.09 ± 0.04 6.58 ± 0.06

Image-to-Image (horse->zebra @ 256x256) Computed using test images

Model Legacy-FID Clean-FID
CycleGAN 77.20 75.17
CUT 45.51 43.71

Building from source

python setup.py bdist_wheel
pip install dist/*

Citation

If you find this repository useful for your research, please cite the following work.

@article{parmar2021cleanfid,
  title={On Buggy Resizing Libraries and Surprising Subtleties in FID Calculation},
  author={Parmar, Gaurav and Zhang, Richard and Zhu, Jun-Yan},
  journal={arXiv preprint arXiv:2104.11222},
  year={2021}
}

Credits

PyTorch-StyleGAN2: code | License

PyTorch-FID: code | License

StyleGAN2: code | LICENSE

converted FFHQ weights: code | License

Using Machine Learning to Test Causal Hypotheses in Conjoint Analysis

Readme File for "Using Machine Learning to Test Causal Hypotheses in Conjoint Analysis" by Ham, Imai, and Janson. (2022) All scripts were written and

0 Jan 27, 2022
This repository contains an implementation of the Permutohedral Attention Module in Pytorch

Permutohedral_attention_module This repository contains an implementation of the Permutohedral Attention Module

Samuel JOUTARD 26 Nov 27, 2022
Automatic Video Captioning Evaluation Metric --- EMScore

Automatic Video Captioning Evaluation Metric --- EMScore Overview For an illustration, EMScore can be computed as: Installation modify the encode_text

Yaya Shi 17 Nov 28, 2022
Pip-package for trajectory benchmarking from "Be your own Benchmark: No-Reference Trajectory Metric on Registered Point Clouds", ECMR'21

Map Metrics for Trajectory Quality Map metrics toolkit provides a set of metrics to quantitatively evaluate trajectory quality via estimating consiste

Mobile Robotics Lab. at Skoltech 31 Oct 28, 2022
ICCV2021 - Mining Contextual Information Beyond Image for Semantic Segmentation

Introduction The official repository for "Mining Contextual Information Beyond Image for Semantic Segmentation". Our full code has been merged into ss

55 Nov 09, 2022
Rapid experimentation and scaling of deep learning models on molecular and crystal graphs.

LitMatter A template for rapid experimentation and scaling deep learning models on molecular and crystal graphs. How to use Clone this repository and

Nathan Frey 32 Dec 06, 2022
QuanTaichi evaluation suite

QuanTaichi: A Compiler for Quantized Simulations (SIGGRAPH 2021) Yuanming Hu, Jiafeng Liu, Xuanda Yang, Mingkuan Xu, Ye Kuang, Weiwei Xu, Qiang Dai, W

Taichi Developers 120 Jan 04, 2023
Web-interface + rest API for classification and regression (https://jeff1evesque.github.io/machine-learning.docs)

Machine Learning This project provides a web-interface, as well as a programmatic-api for various machine learning algorithms. Supported algorithms: S

Jeff Levesque 252 Dec 11, 2022
MixText: Linguistically-Informed Interpolation of Hidden Space for Semi-Supervised Text Classification

MixText This repo contains codes for the following paper: Jiaao Chen, Zichao Yang, Diyi Yang: MixText: Linguistically-Informed Interpolation of Hidden

GT-SALT 309 Dec 12, 2022
An NLP library with Awesome pre-trained Transformer models and easy-to-use interface, supporting wide-range of NLP tasks from research to industrial applications.

简体中文 | English News [2021-10-12] PaddleNLP 2.1版本已发布!新增开箱即用的NLP任务能力、Prompt Tuning应用示例与生成任务的高性能推理! 🎉 更多详细升级信息请查看Release Note。 [2021-08-22]《千言:面向事实一致性的生

6.9k Jan 01, 2023
A Simple LSTM-Based Solution for "Heartbeat Signal Classification and Prediction" in Tianchi

LSTM-Time-Series-Prediction A Simple LSTM-Based Solution for "Heartbeat Signal Classification and Prediction" in Tianchi Contest. The Link of the Cont

KevinCHEN 1 Jun 13, 2022
A standard framework for modelling Deep Learning Models for tabular data

PyTorch Tabular aims to make Deep Learning with Tabular data easy and accessible to real-world cases and research alike.

801 Jan 08, 2023
Single Image Super-Resolution (SISR) with SRResNet, EDSR and SRGAN

Single Image Super-Resolution (SISR) with SRResNet, EDSR and SRGAN Introduction Image super-resolution (SR) is the process of recovering high-resoluti

8 Apr 15, 2022
"SinNeRF: Training Neural Radiance Fields on Complex Scenes from a Single Image", Dejia Xu, Yifan Jiang, Peihao Wang, Zhiwen Fan, Humphrey Shi, Zhangyang Wang

SinNeRF: Training Neural Radiance Fields on Complex Scenes from a Single Image [Paper] [Website] Pipeline Code Environment pip install -r requirements

VITA 250 Jan 05, 2023
Codes for CyGen, the novel generative modeling framework proposed in "On the Generative Utility of Cyclic Conditionals" (NeurIPS-21)

On the Generative Utility of Cyclic Conditionals This repository is the official implementation of "On the Generative Utility of Cyclic Conditionals"

Chang Liu 44 Nov 16, 2022
Face uncertainty quantification or estimation using PyTorch.

Face-uncertainty-pytorch This is a demo code of face uncertainty quantification or estimation using PyTorch. The uncertainty of face recognition is af

Kaen 3 Sep 16, 2022
UFPR-ADMR-v2 Dataset

UFPR-ADMR-v2 Dataset The UFPR-ADMRv2 dataset contains 5,000 dial meter images obtained on-site by employees of the Energy Company of Paraná (Copel), w

Gabriel Salomon 8 Sep 29, 2022
Unofficial implementation of MUSIQ (Multi-Scale Image Quality Transformer)

MUSIQ: Multi-Scale Image Quality Transformer Unofficial pytorch implementation of the paper "MUSIQ: Multi-Scale Image Quality Transformer" (paper link

41 Jan 02, 2023
ESL: Event-based Structured Light

ESL: Event-based Structured Light Video (click on the image) This is the code for the 2021 3DV paper ESL: Event-based Structured Light by Manasi Mugli

Robotics and Perception Group 29 Oct 24, 2022
Minecraft Hack Detection With Python

Minecraft Hack Detection An attempt to try and use crowd sourced replays to find

Kuleen Sasse 3 Mar 26, 2022