按年被提到
20132014 · 4 次2026
视觉、语音与图像生成›卷积网络与图像识别·arXiv:1312.6082 · cs.CV

Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks

用深度卷积神经网络识别街景图像中的多位数字
Ian J. Goodfellow, Yaroslav Bulatov, Julian Ibarz, Sacha Arnoud, Vinay Shet · 2013

单个深度卷积网络直接读街景门牌号,reCAPTCHA 达 99.8%

开始阅读 ↗PDFalphaXiv 讨论
被推荐 · 6 人首次 2014-01-07 · 最近 2017-03-01
6不同的人
6活跃月份
6讨论串
582被引(OpenAlex)

提出一个统一的方法,用一个深度卷积网络直接在像素上同时完成定位、分割和识别,而不是分步处理,并用 DistBelief 训练大规模分布式网络;性能随网络加深而提升,最佳为 11 个隐藏层。在 SVHN 上整串门牌号识别准确率超过 96%,单数字准确率 97.84%;在更难的街景数据上超过 90%;在合成的最难 reCAPTCHA 上达到 99.8%。

HN 为什么提它

评论者主要引用它的 99.8% 这一数字,说明神经网络解 reCAPTCHA 比人更强,也把它当作 Google 街景门牌号识别系统的细节来源。

方法难度 3 / 5约 25 分钟经得起时间
读之前最好知道
1卷积神经网络(CNN)
2深度网络训练与分布式训练
3图像识别基准(SVHN)
4序列输出建模

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
推荐magicalist · 2014-01

“I think the actual paper (linked at the end) is much more informative and interesting than this coverage: Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks”

推荐读原论文,比媒体报道更详尽有趣,讲 Google 街景门牌识别系统
How Google Cracked House Number Identification in Street View原评论 ↗
引用pavelrub · 2014-04

“and it links to the exact same article on arxiv, titled:”

指出该文链接到同名 arXiv 文章,涉及神经网络破解 reCAPTCHA
Google's Street View computer vision can beat reCAPTCHA with 99% accuracy原评论 ↗
引用spcoll · 2014-04

“It seems to be actually 99.8% accuracy according to their paper: Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks”

引用论文指出识别准确率实际为 99.8%
Google's Street View computer vision can beat reCAPTCHA with 99% accuracy原评论 ↗
引用yaroslavvb · 2014-06

“We found that neural networks can solve CAPTCHAS much better than humans, 99.8% on the "hard" ReCAPTCHA instances: Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks”

引用它说明神经网络解 CAPTCHA 比人强,难题上达 99.8%
Captchas are Becoming Ridiculous原评论 ↗
引用cbhl · 2016-02

“99.8% accuracy in solving distorted text reCAPTCHA”

引用它说明扭曲文字 reCAPTCHA 可被 99.8% 的准确率破解
Judge tosses proposed class action accusing Google of CAPTCHA fraud原评论 ↗
引用kastnerkyle · 2016-06

“Sure, there are a lot of potential things to do - either by enumerating towards the full "brute force" loss partway as I think you are hinting toward, or treating as a subsequences, or something else [0, 1, 2].”

把它作为序列标注中构造损失函数思路的参考来源
Intro to Hidden Markov Models (2010) [pdf]原评论 ↗
看全部 7 条 →
引用cbhl · 2017-03

“The state of the art has improved so much that for the older tasks, computers are arguably better than humans at them:”

拿它说明在老任务上计算机可以说已经比人强
ReBreakCaptcha: Breaking Google’s ReCaptcha v2 Using Google原评论 ↗