Official PyTorch Implementation of Fidelity Controllable Extreme Image Compression with GAN

Image Compression using GAN

In this repository, we focus on the compression of images and video (sequence of image frames) using deep generative models and show that they achieve a better performance in compression ratio and perceptual quality. We explore the use of GANs for this task.

Face forgery techniques such as Generative Adversarial Network (GAN) have been widely used for image synthesis in movie production, journalism, etc. What backfires is that these generative technologies are widely abused to impersonate credible people and distribute illegal, misleading, and confusing information to the public. However, to our dismay, the problem with previous fake face detection methods is that they fail to distinguish between different fake generation modalities (various GANs), so none of these methods generalize to opening counterfeit scenes. These previous methods are almost ineffective in identifying fake faces when faced with unknown forgery approaches. To address this challenge, this paper first further analyzes the weaknesses of GAN-based generators. Our validation experimental results of different face generation models, such as Deepfakes, Face2Face, FaceSwap, etc., found that the faces generated by other models have no generalization. Our experiments revealed that the recent fake faces generated by GANs are still not robust enough because it does not consider enough pixels. Inspired by this finding, we design a novel convolutional neural network that uses frequency texture augmentation and knowledge distillation to enhance its global texture perception, effectively describe textures at different semantic levels in images, and improve robustness. It is worth mentioning that we introduce two core components: Discrete Cosine Transform (DCT) and Knowledge Distillation (KDL). DCT could play the role of image compression and also as image distinguishing between fake faces and real faces. KDL is used to extract features from counterfeit and real image targets, making our model generalize to multiple types of fake face generation methods. Experiments were done on two datasets, Celeb-DF and FaceForenscics++, demonstrating that DCT facilitates deep fakes detection in some cases. Knowledge distillation plays a key role in our model. Our model achieves better and more consistent performance in image processing or cross-domain settings, especially when images are subject to Gaussian noise.

deep compression based on the generative models

implementing capsule networks in an image compression GAN

Deep Autoencoder and GAN for image compression

Comparing performance of Vanilla VAE, Info Vae, T Vae and GAN on MNIST image generation and compression

DSA5204 Group13 Project code. A Reproduction of paper CVPR "Learned Image Compression with Mixed Transformer-CNN Architectures" and with extension GAN and CNN with autoencoders

No description available

No description available

No description available

learned image compression using gan

An image compression model using Generative Adversarial Networks (GANs) with low computational cost and very low bitrate

A Generative Advesarial Network Architecture based techniques to compress images and check for the most efficient compression rate

A GANS based image compression method

My Work on Image compression using GAN On different Image Sizes

GAN-based autoencoder for image compression

This repository analyses the SOTA of image compression using Conditional GAN

This project aims to explore the capabilities of advanced deep learning techniques for image compression and generation tasks. Specifically, we are focusing on the integration of autoencoder architectures with Generative Adversarial Networks (GANs) to develop a novel approach for image compression.

This project aims to explore the capabilities of advanced deep learning techniques for image compression and generation tasks. Specifically, we are focusing on the integration of autoencoder architectures with Generative Adversarial Networks (GANs) to develop a novel approach for image compression.

Code and documentations for the project "Compression Reconstruction Loss for Super-Resolution Images with GAN" assigned in the class "Multimedia and Computer Vision" 2020/2021.

GAN / Deepfake Detection — research-oriented ML project for robust real vs synthetic image classification, focused on F1-score and resilience to distribution shifts across generators, compression levels, and data artifacts.

No description available

GAN-Based-Image-Compression

No description available

No description available

Transformer GAN for image compression

No description available

No description available