Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Uncertainty-Aware Audiovisual Activity Recognition Using Deep Bayesian Variational Inference
Mahesh Subedar, Ranganath Krishnan, Paulo Lopez Meyer, Omesh Tickoo, Jonathan Huang
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 stochastic VI Bayesian layers instead of a deterministic DNN head | audiovisual model · MiT subset (in-distribution), audiovisual | top-1 accuracy: 56.61 -> 58.2 | Improved |
| use stochastic VI Bayesian layers instead of a deterministic DNN head | audiovisual model · MiT subset (in-distribution), audiovisual | top-5 accuracy: 79.39 -> 83.8 | Improved |
| use stochastic VI instead of MC dropout | audiovisual Bayesian DNN · MiT subset (in-distribution), audiovisual | top-1 accuracy: 55.04 -> 58.2 | Improved |
| use MC dropout Bayesian inference instead of a deterministic head | audiovisual DNN · MiT subset (in-distribution), audiovisual | top-1 accuracy: 56.61 -> 55.04 | Hurt |
| use MC dropout instead of a deterministic head | audio DNN · MiT subset, audio | top-1 accuracy: 34.13 -> 32.46 | Hurt |
| use stochastic VI instead of a deterministic head | vision DNN · MiT subset, vision | top-1 accuracy: 52.65 -> 53.3 | Improved |
| add audio modality with uncertainty-based fusion to vision | Bayesian DNN (Stochastic VI) · MiT subset | top-1 accuracy: 53.3 -> 58.2 | Improved |
| use stochastic VI instead of a deterministic head | audiovisual model · MiT subset | precision-recall AUC: 10.2% | Improved |
| use stochastic VI instead of MC dropout | audiovisual model · MiT subset | precision-recall AUC: 3.8% | Improved |
| use stochastic VI Bayesian layers instead of a deterministic head | ResNet-101 C3D · UCF101 | top-1 accuracy: 87.5 -> 88.6 | Improved |
Closest papers by meaning
- Bayesian Evidential Deep Learning for Online Action Detection (ECCV 2024 · similarity 0.84)
- Provable Dynamic Fusion for Low-Quality Multimodal Data (ICML 2023 · similarity 0.84)
- A Hierarchical Variational Neural Uncertainty Model for Stochastic Video Prediction (ICCV 2021 · similarity 0.84)
- Bayesian Uncertainty Attribution-Guided Fine-Tuning for Open-Set Action Recognition (ECCV 2026 · similarity 0.84)
- Uncertainty-Boosted Robust Video Activity Anticipation (TPAMI 2024 · similarity 0.84)
- Variational Bayesian Last Layers (ICLR 2024 · similarity 0.82)
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.
