Research map / Probabilistic and scientific modeling / Variational and generative sampling
Unified Multivariate Gaussian Mixture for Efficient Neural Image Compression
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 the vectorized prior, cascaded estimation, and multi-codebook quantization instead of context-enabled baselines | Neural image compression · Kodak, 768 × 512 | whole-compression latency: 3.18× faster than parallel context-enabled models | Helped |
| add cascaded estimation (ablation) | Unified multivariate Gaussian mixture · Kodak | BD-rate: 8.87% -> - | Helped |
| use 2 levels instead of the full 3-level setup (ablation) | Unified multivariate Gaussian mixture · Kodak | BD-rate: 2.33% | Hurt |
| add a fourth estimation level (ablation) | Unified multivariate Gaussian mixture · Kodak | BD-rate: -0.64% | No clear effect |
| replace multi-codebooks with one global codebook (ablation) | Unified multivariate Gaussian mixture · Kodak | BD-rate: 24.40% | Hurt |
| replace Euclidean similarity with cosine similarity (ablation) | Probabilistic vector quantization · Kodak | BD-rate: 4.64% | Hurt |
| replace probabilistic quantization with deterministic quantization (ablation) | Probabilistic vector quantization · Kodak | BD-rate: 16.20%; 11.48% | Hurt |
| add an auxiliary PixelCNN context model | Unified multivariate Gaussian mixture · All levels and groups | bpp: 0.1265 -> 0.1262 | No clear effect |
| add an auxiliary PixelCNN context model | Unified multivariate Gaussian mixture · All levels and groups | context prediction accuracy: 2.24% | No clear effect |
Closest papers by meaning
- Learned Image Compression With Discretized Gaussian Mixture Likelihoods and Attention Modules (CVPR 2020 · similarity 0.90)
- Robustly overfitting latents for flexible neural image compression (NeurIPS 2024 · similarity 0.89)
- Variational image compression with a scale hyperprior (ICLR 2018 · similarity 0.89)
- Improving Inference for Neural Image Compression (NeurIPS 2020 · similarity 0.88)
- Joint Autoregressive and Hierarchical Priors for Learned Image Compression (NeurIPS 2018 · similarity 0.88)
- Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks (NeurIPS 2025 · 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.