Research map / Probabilistic and scientific modeling / Differential equations and PDEs
FVNet: Harnessing Liquid Neural Dynamics for Lightweight Visual Representation
Zhenzhe Hou, Xiaohui Chu, Runze Hu, Yang Li, Yutao Liu
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 FTFU instead of depthwise conv (ablation) | FVNet-T · ImageNet-1K val | top-1 accuracy: 76.1 -> 77.3 | Improved |
| use FTFU instead of MBConv (ablation) | FVNet-T · ImageNet-1K val | top-1 accuracy: 76.8 -> 77.3 | Improved |
| use FTFU instead of LSConv (ablation) | FVNet-T · ImageNet-1K val | top-1 accuracy: 77.0 -> 77.3 | Improved |
| add MPFB to all stages (ablation) | FVNet-T · ImageNet-1K val | top-1 accuracy: 75.8 -> 77.1 (stage3), 77.3 full | Improved |
| add MHSA (ablation) | FVNet-T · ImageNet-1K val | top-1 accuracy: 77.1 -> 77.3 | Improved |
| add GAP gating (ablation) | FVNet-T · ImageNet-1K val | top-1 accuracy: 76.9 -> 77.3 | Improved |
| attention fusion instead of average fusion (ablation) | FVNet-T · ImageNet-1K val | top-1 accuracy: 76.4 -> 77.3 | Improved |
| attention fusion instead of SE-based fusion (ablation) | FVNet-T · ImageNet-1K val | top-1 accuracy: 77.0 -> 77.3 | Improved |
| tanh activation instead of ReLU (ablation) | FVNet-T · ImageNet-1K val | top-1 accuracy: 76.5 -> 77.3 | Improved |
| tanh activation instead of sigmoid (ablation) | FVNet-T · ImageNet-1K val | top-1 accuracy: 76.9 -> 77.3 | Improved |
| FVNet-S vs RepViT-M1.1 on robustness | FVNet-S · ImageNet-C / ImageNet-A / ImageNet-R / ImageNet-Sketch | mCE / top-1: 63.7 -> 62.4 mCE (IN-C) | Improved |
Closest papers by meaning
- V1-Inspired Dynamic Vision System: A Bio-Plausible Video Embedding Framework with Decoupled Shape–Color Pathways and Long-Range Spatiotemporal Perception (NeurIPS 2026 · similarity 0.85)
- CViT: Continuous Vision Transformer for Operator Learning (ICLR 2025 · similarity 0.84)
- Distillation Dynamics: Towards Understanding Feature-Based Distillation in Vision Transformers (AAAI 2026 · similarity 0.84)
- LSNet: See Large, Focus Small (CVPR 2025 · similarity 0.83)
- FasterViT: Fast Vision Transformers with Hierarchical Attention (ICLR 2024 · similarity 0.83)
- VISTREAM: Improving Computation Efficiency of Visual Streaming Perception via Law-of-Charge-Conservation Inspired Spiking Neural Network (CVPR 2025 · similarity 0.83)
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.
