Research map / Image generation / Image restoration
Improved Lossy Image Compression With Priming and Spatially Adaptive Bit Rates for Recurrent Networks
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 |
|---|---|---|---|
| Use DSSIM-weighted loss instead of L1 loss (ablation) | Recurrent compression network · Kodak | MS-SSIM AUC (dB): 32.96 -> 33.43 | Helped |
| Use DSSIM-weighted loss instead of L1 loss (ablation) | Recurrent compression network · Tecnick | MS-SSIM AUC (dB): 35.49 -> 36.02 | Helped |
| Add 3-priming to DSSIM training (ablation) | Recurrent compression network · Kodak | MS-SSIM AUC (dB): 33.43 -> 33.84 | Helped |
| Add 3-priming to DSSIM training (ablation) | Recurrent compression network · Tecnick | MS-SSIM AUC (dB): 36.02 -> 36.34 | Helped |
| Train the 3-priming Prime model for more steps (ablation) | Prime recurrent compression network · Kodak | MS-SSIM AUC (dB): 33.84 -> 34.20 | Helped |
| Train the 3-priming Prime model for more steps (ablation) | Prime recurrent compression network · Tecnick | MS-SSIM AUC (dB): 36.34 -> 36.85 | Helped |
| Increase diffusion from 0 to 1 step (ablation) | DSSIM-trained recurrent compression network · Kodak | MS-SSIM AUC (dB): 31.89 -> 33.05 | Helped |
| Increase diffusion from 1 to 2 steps (ablation) | DSSIM-trained recurrent compression network · Kodak | MS-SSIM AUC (dB): 33.05 -> 32.85 | Hurt |
| Increase diffusion from 2 to 3 steps (ablation) | DSSIM-trained recurrent compression network · Kodak | MS-SSIM AUC (dB): 32.85 -> 33.40 | Helped |
| Increase diffusion from 0 to 1 step (ablation) | DSSIM-trained recurrent compression network · Tecnick | MS-SSIM AUC (dB): 34.34 -> 35.41 | Helped |
| Increase diffusion from 1 to 2 steps (ablation) | DSSIM-trained recurrent compression network · Tecnick | MS-SSIM AUC (dB): 35.41 -> 35.28 | Hurt |
| Increase diffusion from 2 to 3 steps (ablation) | DSSIM-trained recurrent compression network · Tecnick | MS-SSIM AUC (dB): 35.28 -> 35.68 | Helped |
Closest papers by meaning
- Full Resolution Image Compression With Recurrent Neural Networks (CVPR 2017 · similarity 0.92)
- Real-Time Adaptive Image Compression (ICML 2017 · similarity 0.90)
- GLIC: General Format Learned Image Compression (AAAI 2025 · similarity 0.87)
- Towards image compression with perfect realism at ultra-low bitrates (ICLR 2024 · similarity 0.87)
- Towards Backward-Compatible Continual Learning of Image Compression (CVPR 2024 · similarity 0.87)
- Diversify, Contextualize, and Adapt: Efficient Entropy Modeling for Neural Image Codec (NeurIPS 2024 · 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.