Research map / Architectures and efficiency / RNNs and normalization
Learned Image Compression With Discretized Gaussian Mixture Likelihoods and Attention Modules
Check 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 discretized Gaussian mixture likelihoods instead of the Joint entropy model | Learned image compression · Kodim21 | PSNR (dB): 33.435 -> 33.623 | Helped |
| Use discretized Gaussian mixture likelihoods instead of the Joint entropy model | Learned image compression · Kodim21 | MS-SSIM: 0.980 -> 0.981 | Helped |
| Use discretized Gaussian mixture likelihoods instead of the Joint entropy model | Learned image compression · Kodim21 | Rate (bpp): 0.533 -> 0.519 | Helped |
| Add Gaussian mixture likelihoods (ablation) | Learned image compression · Kodak | Rate-distortion performance | Helped |
| Add simplified attention modules (ablation) | Learned image compression · Kodak | Rate-distortion performance | Helped |
| Add attention (ablation) | Learned image compression · Validation loss after 16 epochs | Loss: 3.026 -> 2.754 | Helped |
| Remove the non-local block from the attention module (ablation) | Learned image compression · Validation loss after 16 epochs | Loss: 2.705 -> 2.754 | Hurt |
| Remove the non-local block from the attention module (ablation) | Learned image compression · Training | Time (s)/epoch: 1119 -> 336 | Helped |
| Remove attention (ablation) | Learned image compression · Training | Time (s)/epoch: 336 -> 216 | Helped |
Closest papers by meaning
- Unified Multivariate Gaussian Mixture for Efficient Neural Image Compression (CVPR 2022 · similarity 0.90)
- Towards Efficient Image Compression Without Autoregressive Models (NeurIPS 2023 · similarity 0.90)
- GLIC: General Format Learned Image Compression (AAAI 2025 · similarity 0.89)
- Linear Attention Modeling for Learned Image Compression (CVPR 2025 · similarity 0.89)
- Learned Image Compression with Dictionary-based Entropy Model (CVPR 2025 · similarity 0.89)
- HyperVQ: Enabling Hyperprior Entropy Modeling for VQ-Based Generative Image Compression (NeurIPS 2026 · 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.