Research map / Image generation / Image restoration
Variational image compression with a scale hyperprior
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 learned scale hyperprior to the factorized-prior model | Image compression model · Kodak | rate-distortion performance | Helped |
| Train the hyperprior model using squared-error loss rather than MS-SSIM loss (ablation) | Image compression model · Kodak | PSNR | Helped |
| Train the hyperprior model using MS-SSIM loss rather than squared-error loss (ablation) | Image compression model · Kodak | MS-SSIM | Helped |
| Use MS-SSIM loss instead of squared-error loss (ablation) | Image compression model · Kodak | PSNR | Hurt |
| Use squared-error loss instead of MS-SSIM loss (ablation) | Image compression model · Kodak | MS-SSIM | Mixed |
| Remove batch normalization and learning-rate decay (ablation) | Training setup · Image compression training | compression performance | No clear effect |
| Increase transform filter capacity beyond its saturation point (ablation) | Factorized-prior model · Rate-distortion curves | rate-distortion performance | No clear effect |
Closest papers by meaning
- Improving Inference for Neural Image Compression (NeurIPS 2020 · similarity 0.91)
- End-to-end Optimized Image Compression (ICLR 2017 · similarity 0.90)
- Compressing Images by Encoding Their Latent Representations with Relative Entropy Coding (NeurIPS 2020 · similarity 0.90)
- Robustly overfitting latents for flexible neural image compression (NeurIPS 2024 · similarity 0.89)
- Generative Latent Coding for Ultra-Low Bitrate Image Compression (CVPR 2024 · similarity 0.88)
- Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression (AAAI 2025 · similarity 0.88)
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.