Rethinking the Zigzag Flattening for Image Reading
Not on the map. Why?
Abstract (arXiv)
Sequence ordering of word vector matters a lot to text reading, which has been proven in natural language processing (NLP). However, the rule of different sequence ordering in computer vision (CV) was not well explored, e.g., why the ``zigzag" flattening (ZF) is commonly utilized as a default option to get the image patches ordering in vision networks. Notably, when decomposing multi-scale images, the ZF could not maintain the invariance of feature point positions. To this end, we investigate the Hilbert fractal flattening (HF) as another method for sequence ordering in CV and contrast it against ZF. The HF has proven to be superior to other curves in maintaining spatial locality, when performing multi-scale transformations of dimensional space. And it can be easily plugged into most deep neural networks (DNNs). Extensive experiments demonstrate that it can yield consistent and significant performance boosts for a variety of architectures. Finally, we hope that our studies spark further research about the flattening strategy of image reading.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 31 of its 68 references that are in the map. First posted on arXiv 21 Feb 2022.
Builds on
31 of 68 references.
References · 31
- 2017Attention Is All You Needcited by 59,932
- 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
- 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
- 2021Swin Transformer: Hierarchical Vision Transformer using Shifted Windowscited by 12,082
- 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,893
- 2016Feature Pyramid Networks for Object Detectioncited by 5,021
- 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,581
- 2020Training data-efficient image transformers & distillation through attentioncited by 4,560
- 2016Wide Residual Networkscited by 3,057
- 2021Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutionscited by 1,960
- 2021MLP-Mixer: An all-MLP Architecture for Visioncited by 1,441
- 2018Unified Perceptual Parsing for Scene Understandingcited by 1,265
19 more
- 2021Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gaugescited by 1,131
- 2021Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNetcited by 1,096
- 2021Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolationcited by 1,085
- 2020Designing Network Design Spacescited by 954
- 2021Multiscale Vision Transformerscited by 896
- 2021Do Vision Transformers See Like Convolutional Neural Networks?cited by 671
- 2021Twins: Revisiting the Design of Spatial Attention in Vision Transformerscited by 565
- 2019Res2Net: A New Multi-scale Backbone Architecturecited by 468
- 2014Deeply-Supervised Netscited by 364
- 2021ResMLP: Feedforward networks for image classification with data-efficient trainingcited by 338
- 2021Focal Self-attention for Local-Global Interactions in Vision Transformerscited by 286
- 2020Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depthcited by 241
- 2021Multi-Scale Vision Longformer: A New Vision Transformer for High-Resolution Image Encodingcited by 207
- 2021Rethinking and Improving Relative Position Encoding for Vision Transformercited by 206
- 2021CycleMLP: A MLP-like Architecture for Dense Predictioncited by 126
- 2019Hybrid LSTM and Encoder-Decoder Architecture for Detection of Image Forgeriescited by 59
- 2021Searching the Search Space of Vision Transformercited by 39
- 2021SHAPE: Shifted Absolute Position Embedding for Transformerscited by 37
- 2020Locality Preserving Dense Graph Convolutional Networks with Graph Context-Aware Node Representationscited by 4
Built on by
Citing papers per year: 2025: 1; 2026: 1.
2 papers.