Research map / Image generation / Image restoration
EGIC: Enhanced Low-Bit-Rate Generative Image Compression Guided by Semantic Segmentation
Nikolai Körber, Eduard Kromer, Andreas Siebert, Sascha Hauke, Daniel Mueller-Gritschneder, Björn Schuller
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| replace concat conditioning with pixel-wise projection in OASIS discriminator (ablation) | OASIS · Cityscapes (0.092bpp) | FID: 16.50 -> 13.54 | Improved |
| replace concat conditioning with projection in U-Net discriminator (ablation) | U-Net discriminator · Cityscapes (0.092bpp) | FID: 87.02 -> 30.80 | Improved |
| pre-train the OASIS discriminator with DeepLab2 (ablation) | OASIS · Cityscapes (0.092bpp) | FID: 15.30 -> 9.75 | Improved |
| use weight norm instead of spectral norm (ablation) | OASIS (pre-trained) · Cityscapes (0.092bpp) | FID: 9.75 -> 7.96 | Improved |
| weight norm vs layer norm (ablation) | OASIS discriminator · Cityscapes (0.092bpp) | FID: 9.00 -> 7.96 | Improved |
| add distortion term d to OASIS-C objective (ablation) | OASIS-C · Cityscapes (0.092bpp) | PSNR / FID: 29.97 -> 32.24 PSNR; 7.74 -> 6.36 FID | Improved |
| use stronger segmentation network (DeepLabV3+) as discriminator (ablation) | OASIS · Cityscapes (0.092bpp) | FID: 16.50 -> 20.61 | Hurt |
| add focal frequency loss to pixel-level discriminators (ablation) | OASIS and U-Net · Cityscapes (0.092bpp) | FID: 15.93 vs 16.50 (OASIS) | No clear effect |
| add focal frequency loss to PatchGAN (ablation) | PatchGAN · Cityscapes (0.092bpp) | FID: 112.29 -> 40.64 | Improved |
| class-frequency pixel weighting instead of instance-size weighting (ablation) | OASIS · Cityscapes (0.092bpp) | FID: 15.30 -> 15.56 | No clear effect |
| replace PatchGAN with OASIS-C discriminator (HiFiC backbone, HiFiC-lo) | HiFiC · CLIC 2020 / low bit range | FID: up to 2 FID points, +0.2dB PSNR | Improved |
| replace PatchGAN with OASIS-C discriminator (HiFiC-hi) | HiFiC · high bit-rate | FID/PSNR | Hurt |
Later papers that compared with this method
1 later accepted paper reports EGIC as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- SSCL: Adversarially Guided Image Compression via Semantic and Spectral Consistency Learning (AAAI 2026 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Generative Latent Coding for Ultra-Low Bitrate Image Compression (CVPR 2024 · similarity 0.86)
- Rethinking the Objectives of Vector-Quantized Tokenizers for Image Synthesis (CVPR 2024 · similarity 0.85)
- SeD: Semantic-Aware Discriminator for Image Super-Resolution (CVPR 2024 · similarity 0.85)
- Good, Cheap, and Fast: Overfitted Image Compression with Wasserstein Distortion (CVPR 2025 · similarity 0.85)
- One-Step Diffusion-Based Image Compression with Semantic Distillation (NeurIPS 2025 · similarity 0.85)
- Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression (AAAI 2025 · similarity 0.85)
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.
