Revisiting the Integration of Convolution and Attention for Vision Backbone
- URL: http://arxiv.org/abs/2411.14429v1
- Date: Thu, 21 Nov 2024 18:59:08 GMT
- Title: Revisiting the Integration of Convolution and Attention for Vision Backbone
- Authors: Lei Zhu, Xinjiang Wang, Wayne Zhang, Rynson W. H. Lau,
- Abstract summary: Convolutions and multi-head self-attentions (MHSAs) are typically considered alternatives to each other for building vision backbones.
We propose in this work to use MSHAs and Convs in parallel textbfat different granularity levels instead.
We empirically verify the potential of the proposed integration scheme, named textitGLMix: by offloading the burden of fine-grained features to light-weight Convs, it is sufficient to use MHSAs in a few semantic slots.
- Score: 59.50256661158862
- License:
- Abstract: Convolutions (Convs) and multi-head self-attentions (MHSAs) are typically considered alternatives to each other for building vision backbones. Although some works try to integrate both, they apply the two operators simultaneously at the finest pixel granularity. With Convs responsible for per-pixel feature extraction already, the question is whether we still need to include the heavy MHSAs at such a fine-grained level. In fact, this is the root cause of the scalability issue w.r.t. the input resolution for vision transformers. To address this important problem, we propose in this work to use MSHAs and Convs in parallel \textbf{at different granularity levels} instead. Specifically, in each layer, we use two different ways to represent an image: a fine-grained regular grid and a coarse-grained set of semantic slots. We apply different operations to these two representations: Convs to the grid for local features, and MHSAs to the slots for global features. A pair of fully differentiable soft clustering and dispatching modules is introduced to bridge the grid and set representations, thus enabling local-global fusion. Through extensive experiments on various vision tasks, we empirically verify the potential of the proposed integration scheme, named \textit{GLMix}: by offloading the burden of fine-grained features to light-weight Convs, it is sufficient to use MHSAs in a few (e.g., 64) semantic slots to match the performance of recent state-of-the-art backbones, while being more efficient. Our visualization results also demonstrate that the soft clustering module produces a meaningful semantic grouping effect with only IN1k classification supervision, which may induce better interpretability and inspire new weakly-supervised semantic segmentation approaches. Code will be available at \url{https://github.com/rayleizhu/GLMix}.
Related papers
- MROVSeg: Breaking the Resolution Curse of Vision-Language Models in Open-Vocabulary Semantic Segmentation [33.67313662538398]
We propose a multi-resolution training framework for open-vocabulary semantic segmentation with a single pretrained CLIP backbone.
MROVSeg uses sliding windows to slice the high-resolution input into uniform patches, each matching the input size of the well-trained image encoder.
We demonstrate the superiority of MROVSeg on well-established open-vocabulary semantic segmentation benchmarks.
arXiv Detail & Related papers (2024-08-27T04:45:53Z) - Multi-interactive Feature Learning and a Full-time Multi-modality
Benchmark for Image Fusion and Segmentation [66.15246197473897]
Multi-modality image fusion and segmentation play a vital role in autonomous driving and robotic operation.
We propose a textbfMulti-textbfinteractive textbfFeature learning architecture for image fusion and textbfSegmentation.
arXiv Detail & Related papers (2023-08-04T01:03:58Z) - Image as Set of Points [60.30495338399321]
Context clusters (CoCs) view an image as a set of unorganized points and extract features via simplified clustering algorithm.
Our CoCs are convolution- and attention-free, and only rely on clustering algorithm for spatial interaction.
arXiv Detail & Related papers (2023-03-02T18:56:39Z) - MAFormer: A Transformer Network with Multi-scale Attention Fusion for
Visual Recognition [45.68567088645708]
We introduce Multi-scale Attention Fusion into transformer (MAFormer)
MAFormer explores local aggregation and global feature extraction in a dual-stream framework for visual recognition.
Our MAFormer achieves state-of-the-art performance on common vision tasks.
arXiv Detail & Related papers (2022-08-31T06:29:27Z) - Learning Granularity-Unified Representations for Text-to-Image Person
Re-identification [29.04254233799353]
Text-to-image person re-identification (ReID) aims to search for pedestrian images of an interested identity via textual descriptions.
Existing works usually ignore the difference in feature granularity between the two modalities.
We propose an end-to-end framework based on transformers to learn granularity-unified representations for both modalities, denoted as LGUR.
arXiv Detail & Related papers (2022-07-16T01:26:10Z) - All Grains, One Scheme (AGOS): Learning Multi-grain Instance
Representation for Aerial Scene Classification [31.412401135677744]
We propose a novel all grains, one scheme (AGOS) framework to tackle these challenges.
It consists of a multi-grain perception module (MGP), a multi-branch multi-instance representation module (MBMIR) and a self-aligned semantic fusion (SSF) module.
Our AGOS is flexible and can be easily adapted to existing CNNs in a plug-and-play manner.
arXiv Detail & Related papers (2022-05-06T17:10:44Z) - Vision Transformers with Hierarchical Attention [61.16912607330001]
This paper tackles the high computational/space complexity associated with Multi-Head Self-Attention (MHSA) in vision transformers.
We propose Hierarchical MHSA (H-MHSA), a novel approach that computes self-attention in a hierarchical fashion.
We build a family of Hierarchical-Attention-based Transformer Networks, namely HAT-Net.
arXiv Detail & Related papers (2021-06-06T17:01:13Z) - Conformer: Local Features Coupling Global Representations for Visual
Recognition [72.9550481476101]
We propose a hybrid network structure, termed Conformer, to take advantage of convolutional operations and self-attention mechanisms for enhanced representation learning.
Experiments show that Conformer, under the comparable parameter complexity, outperforms the visual transformer (DeiT-B) by 2.3% on ImageNet.
arXiv Detail & Related papers (2021-05-09T10:00:03Z) - CARAFE++: Unified Content-Aware ReAssembly of FEatures [132.49582482421246]
We propose unified Content-Aware ReAssembly of FEatures (CARAFE++), a universal, lightweight and highly effective operator to fulfill this goal.
CARAFE++ generates adaptive kernels on-the-fly to enable instance-specific content-aware handling.
It shows consistent and substantial gains across all the tasks with negligible computational overhead.
arXiv Detail & Related papers (2020-12-07T07:34:57Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.