Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Scalable Uncertainty for Computer Vision With Functional Variational Inference
Eduardo D. C. Carvalho, Ronald Clark, Andrea Nicastro, Paul H. J. Kelly
Check 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 functional VI with a structured GP variational family and GP prior | FCDenseNet 103 · CamVid | IoU: 0.556 -> 0.623 | Improved |
| use functional VI with a structured GP variational family and GP prior | FCDenseNet 103 · CamVid | Accuracy: 0.893 -> 0.905 | Improved |
| use functional VI with a structured GP variational family and GP prior | FCDenseNet 103 · CamVid test set | Mean Calibration: 0.058 -> 0.053 | Improved |
| replace 50-pass MC dropout with the functional VI prediction method | FCDenseNet 103 · Make3d | Inference time (ms): 2615.65 ± 13.75 -> 50.98 ± 1.74 | Improved |
| replace 50-pass MC dropout with the functional VI prediction method | FCDenseNet 103 · CamVid | Inference time (ms): 5763.63 ± 1.95 -> 128.59 ± 1.86 | Improved |
| use functional VI instead of MC dropout | FCDenseNet 103 · Make3d | rel: 0.210 -> 0.264 | Hurt |
| use functional VI instead of MC dropout | FCDenseNet 103 · Make3d | Mean Calibration: 0.427 -> 0.409 | Improved |
| use berHu likelihood instead of Laplace likelihood (ablation) | Ours regression model · Make3d | Mean Calibration: 0.409 -> 0.631 | Hurt |
| use Gaussian likelihood instead of Laplace likelihood (ablation) | Ours regression model · Make3d | Mean Calibration: 0.409 -> 0.491 | Hurt |
| use berHu likelihood instead of Laplace likelihood (ablation) | Ours regression model · Make3d | rms: 5.74 -> 5.68 | Improved |
Closest papers by meaning
- Beyond the Mean-Field: Structured Deep Gaussian Processes Improve the Predictive Uncertainties (NeurIPS 2020 · similarity 0.90)
- Amortized Structured Stochastic Variational Inference for Gaussian Process Latent Variable Models (NeurIPS 2026 · similarity 0.87)
- VIKING: Deep variational inference with stochastic projections (NeurIPS 2025 · similarity 0.87)
- Bayesian Optimization via Continual Variational Last Layer Training (ICLR 2025 · similarity 0.86)
- Maxitive Donsker-Varadhan Formulation for Possibilistic Variational Inference (NeurIPS 2026 · similarity 0.86)
- Bayesian Basis Function Approximation for Scalable Gaussian Process Priors in Deep Generative Models (ICML 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.
