Research map / Probabilistic and scientific modeling / Variational and generative sampling
FusionVAE: A Deep Hierarchical Variational Autoencoder for RGB Image Fusion
Fabian Duffhauss, Ngo Anh Vien, Hanna Ziesche, Gerhard Neumann
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 posterior q(y) conditioned only on the target instead of q(x,y) (ablation) | FusionVAE · FusionCelebA (avg) | NLL avg: 237.0 -> 233.9 | Improved |
| use posterior q(y) conditioned only on the target instead of q(x,y) (ablation) | FusionVAE · FusionCelebA (avg) | MSE_min avg: 2.04 -> 1.93 | Improved |
| use posterior q(x,y) instead of q(y) (ablation) | FusionVAE · FusionCelebA (3 inputs) | NLL: 228.7 -> 180.5 | Mixed |
| max aggregation instead of Bayesian aggregation (added) (ablation) | FusionVAE · FusionCelebA (avg) | NLL avg: 462.6 -> 233.9 | Improved |
| max aggregation instead of Bayesian aggregation (added) (ablation) | FusionVAE · FusionCelebA (avg) | MSE_min avg: 5.36 -> 1.93 | Improved |
| max aggregation instead of mean aggregation (added) (ablation) | FusionVAE · FusionCelebA (avg) | NLL avg: 231.3 -> 233.9 | No clear effect |
| add encoder features to decoder features instead of aggregating all features (max) (ablation) | FusionVAE · FusionCelebA (avg) | MSE_min avg: 3.03 -> 1.93 | Improved |
| add encoder features to decoder features instead of aggregating all features (max) (ablation) | FusionVAE · FusionCelebA (avg) | NLL avg: 230.6 -> 233.9 | No clear effect |
| add encoder features to decoder features instead of aggregating all features (mean) (ablation) | FusionVAE · FusionCelebA (avg) | MSE_min avg: 2.81 -> 1.98 | Improved |
| use FusionVAE instead of CVAE with skip connections | CVAE+S · FusionCelebA (avg) | NLL avg: 308.3 -> 233.9 | Improved |
| use FusionVAE instead of CVAE | CVAE · FusionMNIST (avg) | MSE_min avg: 1.93 -> 1.45 | Improved |
| use FusionVAE instead of CVAE+S | CVAE+S · FusionMNIST (3 inputs) | NLL: 12.30 -> 13.48 | Hurt |
Closest papers by meaning
- Efficient Rectified Flow for Image Fusion (NeurIPS 2025 · similarity 0.88)
- NVAE: A Deep Hierarchical Variational Autoencoder (NeurIPS 2020 · similarity 0.88)
- VMLoc: Variational Fusion For Learning-Based Multimodal Camera Localization (AAAI 2021 · similarity 0.87)
- A Task-Guided, Implicitly-Searched and Meta-Initialized Deep Model for Image Fusion (TPAMI 2024 · similarity 0.87)
- Hierarchical Decompositional Mixtures of Variational Autoencoders (ICML 2019 · similarity 0.87)
- Multi-Scale Fusion for Object Representation (ICLR 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.
