mlsanityData to 13 Oct 2026

ACNet: Strengthening the Kernel Skeletons for Powerful CNN via Asymmetric Convolution Blocks

arXiv 1908.03930 ↗First posted 11 Aug 2019cs.CV cs.LG cs.NEcited by 178

Not on the map. Why?

Abstract (arXiv)

As designing appropriate Convolutional Neural Network (CNN) architecture in the context of a given application usually involves heavy human works or numerous GPU hours, the research community is soliciting the architecture-neutral CNN structures, which can be easily plugged into multiple mature architectures to improve the performance on our real-world applications. We propose Asymmetric Convolution Block (ACB), an architecture-neutral structure as a CNN building block, which uses 1D asymmetric convolutions to strengthen the square convolution kernels. For an off-the-shelf architecture, we replace the standard square-kernel convolutional layers with ACBs to construct an Asymmetric Convolutional Network (ACNet), which can be trained to reach a higher level of accuracy. After training, we equivalently convert the ACNet into the same original architecture, thus requiring no extra computations anymore. We have observed that ACNet can improve the performance of various models on CIFAR and ImageNet by a clear margin. Through further experiments, we attribute the effectiveness of ACB to its capability of enhancing the model's robustness to rotational distortions and strengthening the central skeleton parts of square convolution kernels.

Key facts. Cited by 178 papers in the map (data through 13 Oct 2026). First posted on arXiv 11 Aug 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2021: 26; 2022: 34; 2023: 39; 2024: 37; 2025: 30; 2026: 12.

178 papers.

Sort:

By area:

  1. 2022YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectorscited by 979
  2. 2021RepVGG: Making VGG-style ConvNets Great Againcited by 520
  3. 2022HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentationcited by 163
  4. 2023FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterizationcited by 161
  5. 2022Scaling Up Your Kernels to 31x31: Revisiting Large Kernel Design in CNNscited by 155
  6. 2022More ConvNets in the 2020s: Scaling up Kernels Beyond 51x51 using Sparsitycited by 142
  7. 2021Salient Object Detection via Integrity Learningcited by 133
  8. 2023RepViT: Revisiting Mobile CNN From ViT Perspectivecited by 123
  9. 2022Self-Sustaining Representation Expansion for Non-Exemplar Class-Incremental Learningcited by 118
  10. 2022MobileOne: An Improved One millisecond Mobile Backbonecited by 109
  11. 2023MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Trainingcited by 96
  12. 2021Diverse Branch Block: Building a Convolution as an Inception-like Unitcited by 87
38 more
  1. 2022What Makes Convolutional Models Great on Long Sequence Modeling?cited by 86
  2. 2023Towards Efficient Visual Adaption via Structural Re-parameterizationcited by 73
  3. 2022Representation Compensation Networks for Continual Semantic Segmentationcited by 72
  4. 2023MotionTrack: Learning Robust Short-term and Long-term Motions for Multi-Object Trackingcited by 64
  5. 2023UniRepLKNet: A Universal Perception Large-Kernel ConvNet for Audio, Video, Point Cloud, Time-Series and Image Recognitioncited by 64
  6. 2021Initialization and Regularization of Factorized Neural Layerscited by 58
  7. 2021Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR-based Perceptioncited by 57
  8. 2024FiLo: Zero-Shot Anomaly Detection by Fine-Grained Description and High-Quality Localizationcited by 51
  9. 2024ClipSAM: CLIP and SAM Collaboration for Zero-Shot Anomaly Segmentationcited by 49
  10. 2024QKFormer: Hierarchical Spiking Transformer using Q-K Attentioncited by 47
  11. 2023NORM: Knowledge Distillation via N-to-One Representation Matchingcited by 44
  12. 2021CTNet: Context-based Tandem Network for Semantic Segmentationcited by 43
  13. 2021RepMLP: Re-parameterizing Convolutions into Fully-connected Layers for Image Recognitioncited by 40
  14. 2022NTIRE 2022 Challenge on Efficient Super-Resolution: Methods and Resultscited by 40
  15. 2024NTIRE 2024 Challenge on Image Super-Resolution (x4): Methods and Resultscited by 40
  16. 2022Overview of Deep Learning-based CSI Feedback in Massive MIMO Systemscited by 37
  17. 2024Improved Implicit Neural Representation with Fourier Reparameterized Trainingcited by 36
  18. 2024Efficient Real-world Image Super-Resolution Via Adaptive Directional Gradient Convolutioncited by 36
  19. 2022Re-parameterizing Your Optimizers rather than Architecturescited by 35
  20. 2021RepMLPNet: Hierarchical Vision MLP with Re-parameterized Localitycited by 32
  21. 2022StructToken : Rethinking Semantic Segmentation with Structural Priorcited by 30
  22. 2023Training Full Spike Neural Networks via Auxiliary Accumulation Pathwaycited by 26
  23. 2025The Tenth NTIRE 2025 Efficient Super-Resolution Challenge Reportcited by 26
  24. 2023VanillaNet: the Power of Minimalism in Deep Learningcited by 23
  25. 2022Make RepVGG Greater Again: A Quantization-aware Approachcited by 22
  26. 2021Collapsible Linear Blocks for Super-Efficient Super Resolutioncited by 19
  27. 2021Real-Time Video Super-Resolution on Smartphones with Deep Learning, Mobile AI 2021 Challenge: Reportcited by 19
  28. 2022RepSR: Training Efficient VGG-style Super-Resolution Networks with Structural Re-Parameterization and Batch Normalizationcited by 19
  29. 2022Edge-enhanced Feature Distillation Network for Efficient Super-Resolutioncited by 18
  30. 2023Consolidator: Mergeable Adapter with Grouped Connections for Visual Adaptationcited by 18
  31. 2021Brain Age Estimation From MRI Using Cascade Networks with Ranking Losscited by 17
  32. 2023AsymFormer: Asymmetrical Cross-Modal Representation Learning for Mobile Platform Real-Time RGB-D Semantic Segmentationcited by 17
  33. 2021Fast Camera Image Denoising on Mobile GPUs with Deep Learning, Mobile AI 2021 Challenge: Reportcited by 15
  34. 2021On-Off Center-Surround Receptive Fields for Accurate and Robust Image Classificationcited by 15
  35. 2024Zero-shot Generalizable Incremental Learning for Vision-Language Object Detectioncited by 15
  36. 2024RingMo-Aerial: An Aerial Remote Sensing Foundation Model With Affine Transformation Contrastive Learningcited by 15
  37. 2024Customize Segment Anything Model for Multi-Modal Semantic Segmentation with Mixture of LoRA Expertscited by 15
  38. 2025LSNet: See Large, Focus Smallcited by 15

50 most cited of 178 All 178 (JSON)