# DyNet: Dynamic Convolution for Accelerating Convolutional Neural Networks

arXiv: https://arxiv.org/abs/2004.10694 · HTML: https://mlsanity.com/paper/2004.10694/

## Key facts

- Cited by 39 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 22 Apr 2020.

## Built on by (most cited first, 30 of 39)

- [Dynamic Spatial Propagation Network for Depth Completion](https://mlsanity.com/paper/2202.09769/) (2022)
- [Decoupled Dynamic Filter Networks](https://mlsanity.com/paper/2104.14107/) (2021)
- [Multi-stage image denoising with the wavelet transform](https://mlsanity.com/paper/2209.12394/) (2022)
- [PointConvFormer: Revenge of the Point-based Convolution](https://mlsanity.com/paper/2208.02879/) (2022)
- [Group R-CNN for Weakly Semi-supervised Object Detection with Points](https://mlsanity.com/paper/2205.05920/) (2022)
- [ELSA: Enhanced Local Self-Attention for Vision Transformer](https://mlsanity.com/paper/2112.12786/) (2021)
- [Lightweight Salient Object Detection in Optical Remote-Sensing Images via Semantic Matching and Edge Alignment](https://mlsanity.com/paper/2301.02778/) (2023)
- [Dynamic MLP for Fine-Grained Image Classification by Leveraging Geographical and Temporal Information](https://mlsanity.com/paper/2203.03253/) (2022)
- [Scale-arbitrary Invertible Image Downscaling](https://mlsanity.com/paper/2201.12576/) (2022)
- [Dynamic Sparse R-CNN](https://mlsanity.com/paper/2205.02101/) (2022)
- [Computational Optics for Mobile Terminals in Mass Production](https://mlsanity.com/paper/2305.05886/) (2023)
- [RefConv: Re-parameterized Refocusing Convolution for Powerful ConvNets](https://mlsanity.com/paper/2310.10563/) (2023)
- [AsConvSR: Fast and Lightweight Super-Resolution Network with Assembled Convolutions](https://mlsanity.com/paper/2305.03387/) (2023)
- [Condition-Adaptive Graph Convolution Learning for Skeleton-Based Gait Recognition](https://mlsanity.com/paper/2308.06707/) (2023)
- [BasisNet: Two-stage Model Synthesis for Efficient Inference](https://mlsanity.com/paper/2105.03014/) (2021)
- [TVConv: Efficient Translation Variant Convolution for Layout-aware Visual Processing](https://mlsanity.com/paper/2203.10489/) (2022)
- [FalconNet: Factorization for the Light-weight ConvNets](https://mlsanity.com/paper/2306.06365/) (2023)
- [Collaboration of Experts: Achieving 80% Top-1 Accuracy on ImageNet with 100M FLOPs](https://mlsanity.com/paper/2107.03815/) (2021)
- [Deeper Insights into the Robustness of ViTs towards Common Corruptions](https://mlsanity.com/paper/2204.12143/) (2022)
- [Tunable Convolutions with Parametric Multi-Loss Optimization](https://mlsanity.com/paper/2304.00898/) (2023)
- [DRPN: Making CNN Dynamically Handle Scale Variation](https://mlsanity.com/paper/2112.10963/) (2021)
- [RepMode: Learning to Re-parameterize Diverse Experts for Subcellular Structure Prediction](https://mlsanity.com/paper/2212.10066/) (2022)
- [Dynamic Convolutional Neural Networks as Efficient Pre-trained Audio Models](https://mlsanity.com/paper/2310.15648/) (2023)
- [Image Captioning via Dynamic Path Customization](https://mlsanity.com/paper/2406.00334/) (2024)
- [Towards Understanding the Effectiveness of Attention Mechanism](https://mlsanity.com/paper/2106.15067/) (2021)
- [Evolution: A Unified Formula for Feature Operators from a High-level Perspective](https://mlsanity.com/paper/2305.14409/) (2023)
- [Multi-Convformer: Extending Conformer with Multiple Convolution Kernels](https://mlsanity.com/paper/2407.03718/) (2024)
- [Instance and Pair-Aware Dynamic Networks for Re-Identification](https://mlsanity.com/paper/2103.05395/) (2021)
- [LAConv: Local Adaptive Convolution for Image Fusion](https://mlsanity.com/paper/2107.11617/) (2021)
- [Adaptive Blind Super-Resolution Network for Spatial-Specific and Spatial-Agnostic Degradations](https://mlsanity.com/paper/2506.07705/) (2025)

## Abstract (arXiv)

Convolution operator is the core of convolutional neural networks (CNNs) and occupies the most computation cost. To make CNNs more efficient, many methods have been proposed to either design lightweight networks or compress models. Although some efficient network structures have been proposed, such as MobileNet or ShuffleNet, we find that there still exists redundant information between convolution kernels. To address this issue, we propose a novel dynamic convolution method to adaptively generate convolution kernels based on image contents. To demonstrate the effectiveness, we apply dynamic convolution on multiple state-of-the-art CNNs. On one hand, we can reduce the computation cost remarkably while maintaining the performance. For ShuffleNetV2/MobileNetV2/ResNet18/ResNet50, DyNet can reduce 37.0/54.7/67.2/71.3% FLOPs without loss of accuracy. On the other hand, the performance can be largely boosted if the computation cost is maintained. Based on the architecture MobileNetV3-Small/Large, DyNet achieves 70.3/77.1% Top-1 accuracy on ImageNet with an improvement of 2.9/1.9%. To verify the scalability, we also apply DyNet on segmentation task, the results show that DyNet can reduce 69.3% FLOPs while maintaining Mean IoU on segmentation task.