Research map / Image generation / Autoregressive and latent generation
Learning Diverse Image Colorization
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 a VAE color-field embedding with an MDN conditional model rather than CVAE | Diverse image colorization · LFW | Var.: $1.0 \times10^{-4} -> $1.1 \times10^{-3} | Helped |
| Use a VAE color-field embedding with an MDN conditional model rather than cGAN | Diverse image colorization · LFW | Var.: $8.4 \times10^{-6} -> $1.1 \times10^{-3} | Helped |
| Add a grey-level encoder and skip connections (ablation) | Ours · LFW | Var.: $1.1 \times10^{-3} -> $4.4 \times10^{-4} | Hurt |
| Add a grey-level encoder and skip connections (ablation) | Ours · ImageNet-Val | Eob.: .043 -> .041 | Helped |
| Use all proposed decoder terms rather than standard L2 loss (ablation) | VAE decoder · LFW | mean absolute error per pixel (All): .034 -> .029 | Helped |
| Use all proposed decoder terms rather than standard L2 loss (ablation) | VAE decoder · LSUN Church | mean absolute error per pixel (All): .024 -> .023 | Helped |
| Use all proposed decoder terms rather than standard L2 loss (ablation) | VAE decoder · ImageNet-Val | mean absolute error per pixel (All): {f .031} -> .039 | Hurt |
| Use all proposed decoder terms rather than standard L2 loss (ablation) | VAE decoder · LFW | weighted absolute error per pixel (All): 7.20 -> 2.65 | Helped |
| Use all proposed decoder terms rather than standard L2 loss (ablation) | VAE decoder · LSUN Church | weighted absolute error per pixel (All): 4.9 -> 1.74 | Helped |
| Use all proposed decoder terms rather than standard L2 loss (ablation) | VAE decoder · ImageNet-Val | weighted absolute error per pixel (All): 10.02 -> 4.82 | Helped |
| Add colorfulness and gradient losses to specificity loss (ablation) | VAE decoder · LFW | weighted absolute error per pixel (All): 6.69 -> 2.65 | Helped |
| Add colorfulness and gradient losses to specificity loss (ablation) | VAE decoder · LSUN Church | weighted absolute error per pixel (All): 6.54 -> 1.74 | Helped |
Closest papers by meaning
- Colorization Transformer (ICLR 2021 · similarity 0.89)
- Diversity-Sensitive Conditional Generative Adversarial Networks (ICLR 2019 · similarity 0.87)
- Instance-Aware Image Colorization (CVPR 2020 · similarity 0.87)
- Discovering Latent Graphs with GFlowNets for Diverse Conditional Image Generation (NeurIPS 2025 · similarity 0.85)
- Automatic Controllable Colorization via Imagination (CVPR 2024 · similarity 0.84)
- Color-Oriented Redundancy Reduction in Dataset Distillation (NeurIPS 2024 · similarity 0.84)
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.