Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/zohebabai/tf-focal-frequency-loss
Tensorflow Implementation of Focal Frequency Loss for Image Reconstruction and Synthesis [ICCV 2021]
https://github.com/zohebabai/tf-focal-frequency-loss
autoencoder frequency-analysis frequency-domain gan generative-adversarial-network generative-models generic iccv2021 image-generation image-reconstruction image-synthesis loss loss-function pix2pix spade stylegan2 variational-autoencoder
Last synced: about 2 months ago
JSON representation
Tensorflow Implementation of Focal Frequency Loss for Image Reconstruction and Synthesis [ICCV 2021]
- Host: GitHub
- URL: https://github.com/zohebabai/tf-focal-frequency-loss
- Owner: ZohebAbai
- License: mit
- Created: 2022-05-30T06:57:50.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2022-05-30T14:53:31.000Z (over 2 years ago)
- Last Synced: 2024-12-03T14:23:35.917Z (2 months ago)
- Topics: autoencoder, frequency-analysis, frequency-domain, gan, generative-adversarial-network, generative-models, generic, iccv2021, image-generation, image-reconstruction, image-synthesis, loss, loss-function, pix2pix, spade, stylegan2, variational-autoencoder
- Language: Python
- Homepage:
- Size: 10.7 KB
- Stars: 9
- Watchers: 3
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Focal Frequency Loss - Tensorflow Implementation
[](https://badge.fury.io/py/tf-focal-frequency-loss)
This repository provides the Tensorflow implementation for the following paper:
**Focal Frequency Loss for Image Reconstruction and Synthesis** by Liming Jiang, Bo Dai, Wayne Wu and Chen Change Loy
in ICCV 2021.
[**Project Page**](https://www.mmlab-ntu.com/project/ffl/index.html) | [**Paper**](https://arxiv.org/abs/2012.12821) | [**Poster**](https://liming-jiang.com/projects/FFL/resources/poster.pdf) | [**Slides**](https://liming-jiang.com/projects/FFL/resources/slides.pdf) | [**YouTube Demo**](https://www.youtube.com/watch?v=RNTnDtKvcpc) | [**Official PyTorch Implementation**](https://github.com/EndlessSora/focal-frequency-loss)
> **Abstract:** *Image reconstruction and synthesis have witnessed remarkable progress thanks to the development of generative models. Nonetheless, gaps could still exist between the real and generated images, especially in the frequency domain. In this study, we show that narrowing gaps in the frequency domain can ameliorate image reconstruction and synthesis quality further. We propose a novel focal frequency loss, which allows a model to adaptively focus on frequency components that are hard to synthesize by down-weighting the easy ones. This objective function is complementary to existing spatial losses, offering great impedance against the loss of important frequency information due to the inherent bias of neural networks. We demonstrate the versatility and effectiveness of focal frequency loss to improve popular models, such as VAE, pix2pix, and SPADE, in both perceptual quality and quantitative performance. We further show its potential on StyleGAN2.*
## Quick Start
Run `pip install tf-focal-frequency-loss` for installation. Then, the following code is all you need.
```python
import tensorflow as tf
from tf_focal_frequency_loss import FocalFrequencyLoss as FFL
ffl = FFL(loss_weight=1.0, alpha=1.0) # initialize tf.keras.layers.Layer class
fake = tf.random.normal((4, 3, 64, 64)) # replace it with the predicted tensor of shape (N, C, H, W)
real = tf.random.normal((4, 3, 64, 64)) # replace it with the target tensor of shape (N, C, H, W)
loss = ffl(fake, real) # calculate focal frequency loss
```
## Notebook
[**Pix2Pix**](https://github.com/ZohebAbai/Deep-Learning-Projects/blob/master/08_TF_Pix2Pix_on_Edges2Handbags.ipynb)
## License
All rights reserved. The code is released under the [MIT License](https://github.com/ZohebAbai/tf-focal-frequency-loss/blob/main/LICENSE).