Research map / Image generation / Image restoration
Flexible Neural Image Compression via Code Editing
Chenjian Gao, Tongda Xu, Dailan He, Yan Wang, Hongwei Qin
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 |
|---|---|---|---|
| add adaptive quantization step Delta (Enhanced vs Naive) (ablation) | Code Editing · Kodak | R-D performance | Improved |
| control Delta alone without optimizing y (ablation) | Code Editing · Kodak | R-D performance | Hurt |
| use SGA instead of additive uniform noise relaxation (ablation) | Code Editing · Kodak | R-D performance | Improved |
| AUN relaxation vs baseline without code editing (ablation) | Code Editing · Kodak | R-D performance | Improved |
| terminate Code Editing early with 50/100/200 iterations (ablation) | Code Editing · Kodak | R-D performance: {50, 100, 200} | Mixed |
| fine-tune encoder with lambda=0.075 to initialize Code Editing (ablation) | Code Editing · Kodak high bitrate | R-D performance | Improved |
| Code Editing Enhanced vs existing variable bitrate methods | Code Editing Enhanced · Kodak | R-D performance | Improved |
| increase LPIPS weight lambda_p from 0.1 to 1.0 | Code Editing Enhanced · Kodak | PSNR and LPIPS: {0.1, 0.5, 1.0} | Mixed |
| optimize Delta_z by gradient instead of grid search (ablation) | Code Editing Enhanced · Kodak | R-D performance | Hurt |
| fix Delta_z=1.0 with no grid search (ablation) | Code Editing Enhanced · Kodak | R-D performance | No clear effect |
| compare to SGA applied per baseline model (multiple decoders) | Code Editing Enhanced · Kodak | R-D performance | Mixed |
| extend bitrate beyond 1.0 bpp from decoder trained at 0.5 bpp (ablation) | Code Editing Enhanced · Kodak | R-D performance | Hurt |
Closest papers by meaning
- Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression (AAAI 2025 · similarity 0.88)
- Neural Video Compression with Feature Modulation (CVPR 2024 · similarity 0.87)
- On Quantizing Neural Representation for Variable-Rate Video Coding (ICLR 2025 · similarity 0.86)
- Robustly overfitting latents for flexible neural image compression (NeurIPS 2024 · similarity 0.86)
- Towards Backward-Compatible Continual Learning of Image Compression (CVPR 2024 · similarity 0.86)
- Towards Practical Real-Time Neural Video Compression (CVPR 2025 · similarity 0.86)
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.
