Focal Modulation Networks

计算机科学 分割 人工智能 安全性令牌 卷积神经网络 推论 模式识别(心理学) 仿射变换 计算机视觉 计算机安全 纯数学 数学
作者
Jianwei Yang,Chunyuan Li,Jianfeng Gao
出处
期刊:Cornell University - arXiv 被引量:116
标识
DOI:10.48550/arxiv.2203.11926
摘要

We propose focal modulation networks (FocalNets in short), where self-attention (SA) is completely replaced by a focal modulation mechanism for modeling token interactions in vision. Focal modulation comprises three components: (i) hierarchical contextualization, implemented using a stack of depth-wise convolutional layers, to encode visual contexts from short to long ranges, (ii) gated aggregation to selectively gather contexts for each query token based on its content, and (iii) element-wise modulation or affine transformation to inject the aggregated context into the query. Extensive experiments show FocalNets outperform the state-of-the-art SA counterparts (e.g., Swin and Focal Transformers) with similar computational costs on the tasks of image classification, object detection, and segmentation. Specifically, FocalNets with tiny and base size achieve 82.3% and 83.9% top-1 accuracy on ImageNet-1K. After pretrained on ImageNet-22K in 224 resolution, it attains 86.5% and 87.3% top-1 accuracy when finetuned with resolution 224 and 384, respectively. When transferred to downstream tasks, FocalNets exhibit clear superiority. For object detection with Mask R-CNN, FocalNet base trained with 1\times outperforms the Swin counterpart by 2.1 points and already surpasses Swin trained with 3\times schedule (49.0 v.s. 48.5). For semantic segmentation with UPerNet, FocalNet base at single-scale outperforms Swin by 2.4, and beats Swin at multi-scale (50.5 v.s. 49.7). Using large FocalNet and Mask2former, we achieve 58.5 mIoU for ADE20K semantic segmentation, and 57.9 PQ for COCO Panoptic Segmentation. Using huge FocalNet and DINO, we achieved 64.3 and 64.4 mAP on COCO minival and test-dev, respectively, establishing new SoTA on top of much larger attention-based models like Swinv2-G and BEIT-3. Code and checkpoints are available at https://github.com/microsoft/FocalNet.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
kk发布了新的文献求助10
刚刚
66完成签到 ,获得积分10
刚刚
111完成签到 ,获得积分10
刚刚
happyou完成签到,获得积分10
刚刚
彩色半烟完成签到,获得积分10
1秒前
Everything发布了新的文献求助10
1秒前
1秒前
OOK发布了新的文献求助10
1秒前
are完成签到,获得积分10
1秒前
我想打喷嚏完成签到,获得积分10
1秒前
瀚泛完成签到,获得积分10
2秒前
共享精神应助尛瞐慶成采纳,获得10
3秒前
ybr完成签到,获得积分20
4秒前
happyou发布了新的文献求助10
4秒前
饱满友易发布了新的文献求助10
5秒前
无机盐完成签到,获得积分10
6秒前
manny发布了新的文献求助10
6秒前
老迟到的翠容完成签到,获得积分10
6秒前
ark861023完成签到,获得积分10
7秒前
m123完成签到,获得积分10
7秒前
had完成签到,获得积分10
8秒前
8秒前
FashionBoy应助冷眼观潮采纳,获得10
8秒前
FF发布了新的文献求助10
8秒前
不穷知识发布了新的文献求助10
8秒前
8秒前
8秒前
Minguk完成签到,获得积分10
9秒前
小白完成签到,获得积分10
9秒前
志士心完成签到,获得积分10
10秒前
10秒前
大大大大管子完成签到 ,获得积分10
11秒前
顾矜应助Everything采纳,获得10
11秒前
伟大的鲁路皇完成签到,获得积分10
11秒前
zzg发布了新的文献求助10
13秒前
王静姝完成签到,获得积分10
13秒前
Akim应助神勇的天问采纳,获得10
13秒前
时舒完成签到 ,获得积分10
14秒前
14秒前
14秒前
高分求助中
Mass producing individuality 600
Algorithmic Mathematics in Machine Learning 500
Разработка метода ускоренного контроля качества электрохромных устройств 500
Advances in Underwater Acoustics, Structural Acoustics, and Computational Methodologies 300
A Student's Guide to Maxwell's Equations 200
The Power of High-Throughput Experimentation: General Topics and Enabling Technologies for Synthesis and Catalysis (Volume 1) 200
NK Cell Receptors: Advances in Cell Biology and Immunology by Colton Williams (Editor) 200
热门求助领域 (近24小时)
化学 材料科学 医学 生物 工程类 有机化学 物理 生物化学 纳米技术 计算机科学 化学工程 内科学 复合材料 物理化学 电极 遗传学 量子力学 基因 冶金 催化作用
热门帖子
关注 科研通微信公众号,转发送积分 3827474
求助须知:如何正确求助?哪些是违规求助? 3369741
关于积分的说明 10457440
捐赠科研通 3089439
什么是DOI,文献DOI怎么找? 1699861
邀请新用户注册赠送积分活动 817560
科研通“疑难数据库(出版商)”最低求助积分说明 770263