From Global to Local: Multi-Scale Out-of-Distribution Detection

判别式 计算机科学 成对比较 人工智能 模式识别(心理学) 代表(政治) 杂乱 机器学习 电信 雷达 政治 政治学 法学
作者
Ji Zhang,Lianli Gao,Bingguang Hao,Hao Huang,Jingkuan Song,Heng Tao Shen
出处
期刊:IEEE transactions on image processing [Institute of Electrical and Electronics Engineers]
卷期号:32: 6115-6128 被引量:10
标识
DOI:10.1109/tip.2023.3328478
摘要

Out-of-distribution (OOD) detection aims to detect "unknown" data whose labels have not been seen during the in-distribution (ID) training process. Recent progress in representation learning gives rise to distance-based OOD detection that recognizes inputs as ID/OOD according to their relative distances to the training data of ID classes. Previous approaches calculate pairwise distances relying only on global image representations, which can be sub-optimal as the inevitable background clutter and intra-class variation may drive image-level representations from the same ID class far apart in a given representation space. In this work, we overcome this challenge by proposing Multi-scale OOD DEtection (MODE), a first framework leveraging both global visual information and local region details of images to maximally benefit OOD detection. Specifically, we first find that existing models pretrained by off-the-shelf cross-entropy or contrastive losses are incompetent to capture valuable local representations for MODE, due to the scale-discrepancy between the ID training and OOD detection processes. To mitigate this issue and encourage locally discriminative representations in ID training, we propose Attention-based Local PropAgation (ALPA), a trainable objective that exploits a cross-attention mechanism to align and highlight the local regions of the target objects for pairwise examples. During test-time OOD detection, a Cross-Scale Decision (CSD) function is further devised on the most discriminative multi-scale representations to distinguish ID/OOD data more faithfully. We demonstrate the effectiveness and flexibility of MODE on several benchmarks - on average, MODE outperforms the previous state-of-the-art by up to 19.24% in FPR, 2.77% in AUROC. Code is available at https://github.com/JimZAI/MODE-OOD.
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
更新
大幅提高文件上传限制,最高150M (2024-4-1)

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
李爱国应助Kenzonvay采纳,获得10
1秒前
hw发布了新的文献求助10
1秒前
冰冰发布了新的文献求助10
1秒前
1秒前
Sakura发布了新的文献求助10
1秒前
星辰大海应助小溪采纳,获得10
1秒前
Wei Qin发布了新的文献求助10
2秒前
YT应助strugglejsp采纳,获得10
2秒前
2秒前
木马病毒完成签到 ,获得积分10
2秒前
3秒前
心灵美复天完成签到,获得积分10
3秒前
bkagyin应助攒星星采纳,获得10
4秒前
深情安青应助Nancy采纳,获得10
5秒前
Once完成签到,获得积分10
5秒前
Nagi完成签到,获得积分10
5秒前
123发布了新的文献求助10
6秒前
科研通AI2S应助欣欣丽丽采纳,获得10
7秒前
7秒前
852应助禹无极采纳,获得10
8秒前
8秒前
8秒前
9秒前
9秒前
Wei Qin完成签到,获得积分10
10秒前
鱼儿123发布了新的文献求助30
10秒前
Olivia发布了新的文献求助10
11秒前
gggja发布了新的文献求助10
11秒前
12秒前
Ooops完成签到,获得积分10
12秒前
12秒前
生工跑路ing完成签到,获得积分10
12秒前
小蘑菇应助敬老院N号采纳,获得10
12秒前
13秒前
13秒前
13秒前
14秒前
呆呆是一条鱼完成签到,获得积分10
14秒前
14秒前
cctv18应助mysoul123采纳,获得10
14秒前
高分求助中
좌파는 어떻게 좌파가 됐나:한국 급진노동운동의 형성과 궤적 2500
Sustainability in Tides Chemistry 1500
TM 5-855-1(Fundamentals of protective design for conventional weapons) 1000
CLSI EP47 Evaluation of Reagent Carryover Effects on Test Results, 1st Edition 800
Threaded Harmony: A Sustainable Approach to Fashion 799
Livre et militantisme : La Cité éditeur 1958-1967 500
Retention of title in secured transactions law from a creditor's perspective: A comparative analysis of selected (non-)functional approaches 500
热门求助领域 (近24小时)
化学 医学 生物 材料科学 工程类 有机化学 生物化学 物理 内科学 纳米技术 计算机科学 化学工程 复合材料 基因 遗传学 催化作用 物理化学 免疫学 量子力学 细胞生物学
热门帖子
关注 科研通微信公众号,转发送积分 3054545
求助须知:如何正确求助?哪些是违规求助? 2711512
关于积分的说明 7426610
捐赠科研通 2356104
什么是DOI,文献DOI怎么找? 1247642
科研通“疑难数据库(出版商)”最低求助积分说明 606478
版权声明 596079