Research map / Video / Action recognition
Volterra Neural Networks (VNNs)
Siddharth Roheda, Hamid Krim, Bo Jiang
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 |
|---|---|---|---|
| fuse RGB and optical flow with Volterra filter instead of feature concatenation (ablation) | O-VNN-H from scratch · UCF-101 | accuracy: 82.31 -> 90.28 | Improved |
| fuse with Volterra filter instead of weighted averaging (ablation) | O-VNN-H from scratch · UCF-101 | accuracy: 85.79 -> 90.28 | Improved |
| fuse with Volterra filter instead of event driven fusion (ablation) | O-VNN-H from scratch · HMDB-51 | accuracy: 60.36 -> 65.61 | Improved |
| add cascaded structure and higher complexity to plain Volterra filter (ablation) | O-VNN-H RGB only from scratch · UCF-101 | accuracy: 38.19 -> 58.73 | Improved |
| use O-VNN-H instead of lighter O-VNN-L (ablation) | two-stream from scratch · UCF-101 | accuracy: 86.16 -> 90.28 | Improved |
| add second stream (optical flow) to RGB-only O-VNN-H | O-VNN-H from scratch · UCF-101 | accuracy: 58.73 -> 90.28 | Improved |
| pretrain O-VNN-H on Kinetics | two-stream O-VNN-H · UCF-101 | accuracy: 90.28 -> 98.49 | Improved |
| replace conv layers with Volterra layers in two-stream network | Kinetics-400 · Kinetics-400 | accuracy: 74.2 -> 78.8 | Improved |
| use Volterra generator in WGAN | WGAN · CIFAR10 | inception score: 5.95 -> 6.20 | Improved |
| use Volterra layers with fixed dictionary in STGAN | STGAN · CIFAR10 | inception score: 6.80 -> 7.05 | Improved |
| increase number of multipliers Q in cascaded Volterra filter (ablation) | cascaded Volterra filter · training loss | training loss | Improved |
Closest papers by meaning
- Conquering the CNN Over-Parameterization Dilemma: A Volterra Filtering Approach for Action Recognition (AAAI 2020 · similarity 0.98)
- Toward Understanding Convolutional Neural Networks from Volterra Convolution Perspective (JMLR 2022 · similarity 0.88)
- kVNN: Learnable Volterra Network Kernels (NeurIPS 2026 · similarity 0.87)
- MR-VNet: Media Restoration using Volterra Networks (CVPR 2024 · similarity 0.87)
- Dynamic Filter Networks (NeurIPS 2016 · similarity 0.85)
- AsCAN: Asymmetric Convolution-Attention Networks for Efficient Recognition and Generation (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.
