Research map / Image generation / Image restoration
High-Fidelity Generative Image Compression
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add a GAN loss to the MSE+LPIPS compression objective (ablation) | HiFiC · CLIC2020 testset | perceptual quality indices | Helped |
| Add a GAN loss to the MSE+LPIPS compression objective (ablation) | HiFiC · CLIC2020 testset | PSNR | Hurt |
| Add a GAN loss to the MSE+LPIPS compression objective (ablation) | HiFiC · CLIC2020 user study | human preference | Helped |
| Condition the discriminator on the latent y (ablation) | HiFiC discriminator · CLIC2020 testset | perceptual quality indices | Helped |
| Condition the discriminator on the latent y (ablation) | HiFiC discriminator · CLIC2020 testset | FID run-to-run variability | Helped |
| Use SpectralNorm instead of InstanceNorm in a non-conditional discriminator (ablation) | HiFiC discriminator · CLIC2020 testset | FID run-to-run variability: 3.4 -> 0.14 | Helped |
| Use ChannelNorm instead of InstanceNorm in the encoder and generator (ablation) | HiFiC at image resolutions different from training crops · High-resolution image evaluation | resolution-generalization artifacts | Helped |
| Use BatchNorm in the encoder and generator (ablation) | HiFiC · Training | training stability | Hurt |
| Increase generator residual blocks from 1 toward 5 (ablation) | HiFiC generator · CLIC2020 testset | FID | Helped |
| Increase generator residual blocks from 1 toward 5 (ablation) | HiFiC generator · CLIC2020 testset | PSNR | Helped |
| Remove MSE and LPIPS distortion losses while retaining GAN training and learn the encoder (ablation) | HiFiC · Training | GAN training stability | Hurt |
| Add LPIPS rather than omit perceptual distortion d_P (ablation) | HiFiC · Image reconstruction | gridding artifacts | Helped |
Closest papers by meaning
- How To Exploit the Transferability of Learned Image Compression to Conventional Codecs (CVPR 2021 · similarity 0.92)
- Towards image compression with perfect realism at ultra-low bitrates (ICLR 2024 · similarity 0.89)
- Good, Cheap, and Fast: Overfitted Image Compression with Wasserstein Distortion (CVPR 2025 · similarity 0.89)
- Generative Latent Coding for Ultra-Low Bitrate Image Compression (CVPR 2024 · similarity 0.89)
- Once-for-All: Controllable Generative Image Compression with Dynamic Granularity Adaptation (ICLR 2025 · similarity 0.87)
- Decouple Distortion from Perception: Region Adaptive Diffusion for Extreme-low Bitrate Perception Image Compression (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.