Research map / Image generation / Image restoration
Lossy Image Compression with Compressive Autoencoders
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 rounding in the forward pass and the identity derivative in the backward pass | Compressive autoencoder · Kodak PhotoCD | SSIM | Helped |
| Use rounding in the forward pass and the identity derivative in the backward pass | Compressive autoencoder · Kodak PhotoCD MOS test | mean opinion score | Helped |
| Use rounding in the forward pass and the identity derivative in the backward pass | Compressive autoencoder · Kodak PhotoCD | PSNR | Mixed |
| Use rounding in the forward pass and the identity derivative in the backward pass | Compressive autoencoder · Kodak PhotoCD | MS-SSIM | No clear effect |
| Replace rounding-based training with additive uniform noise (ablation) | Compressive autoencoder · High-bit-rate compression | rate-distortion performance | Hurt |
| Train coefficients incrementally by enabling them through a binary mask (ablation) | Compressive autoencoder · Training | training stability and performance | Helped |
| Train all coefficients from the start instead of enabling them incrementally (ablation) | Compressive autoencoder · Training | training stability and performance | Hurt |
| Use a small learning rate from the beginning for non-incremental training (ablation) | Compressive autoencoder · Training | training speed | Hurt |
| Remove quantization while retaining the encoder's dimensionality reduction (ablation) | Compressive autoencoder · Image reconstruction | visible reconstruction artifacts | Helped |
| Optimize all network parameters for a particular rate-distortion tradeoff instead of fine-tuning one model across rates (ablation) | Compressive autoencoder · Natural-image compression across bit rates | distortion | Helped |
Closest papers by meaning
- Variational image compression with a scale hyperprior (ICLR 2018 · similarity 0.90)
- Improved Lossy Image Compression With Priming and Spatially Adaptive Bit Rates for Recurrent Networks (CVPR 2018 · similarity 0.90)
- Compression of Structured Data with Autoencoders: Provable Benefit of Nonlinearities and Depth (ICML 2024 · similarity 0.88)
- Robustly overfitting latents for flexible neural image compression (NeurIPS 2024 · similarity 0.86)
- DC-SAE: Deep Compression Semantic Autoencoder for Faster Diffusion Convergence (NeurIPS 2026 · similarity 0.86)
- Auto-Compressing Networks (NeurIPS 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.