【摘要】：2006年深度學(xué)習(xí)的提出為機(jī)器學(xué)習(xí)領(lǐng)域帶來新的革命,深度學(xué)習(xí)的成功不僅依賴于理論和模型上的突破,也離不開大數(shù)據(jù)環(huán)境下的海量訓(xùn)練樣本以及不斷革新的先進(jìn)計(jì)算技術(shù)。在GPU被應(yīng)用于科學(xué)計(jì)算之前,神經(jīng)網(wǎng)絡(luò)特別是大型神經(jīng)網(wǎng)絡(luò)的訓(xùn)練時間往往令人生畏。GPU特別適應(yīng)于并行計(jì)算的特性給神經(jīng)網(wǎng)絡(luò)的訓(xùn)練速度帶來數(shù)十倍的提升。開源的GPU計(jì)算框架也不斷地推陳出新,推動深度學(xué)習(xí)在各方面的應(yīng)用,Caffe就是其中的一種。由于簡單易用、性能強(qiáng)大,Caffe框架受到了廣泛的認(rèn)可。利用Caffe框架對印章類型進(jìn)行識別,所采用的兩種模型都取得極好的實(shí)驗(yàn)效果,對印章的自動識別提供新的參考。
[Abstract]:In 2006, the proposal of deep learning brought a new revolution to the field of machine learning. The success of deep learning not only depends on the breakthrough of theory and model, but also depends on the massive training samples and innovative advanced computing technology in big data environment. Before GPU was applied to scientific computing, the training time of neural network, especially large neural network, was often daunting. GPU is especially suitable for parallel computing, which brings dozens of times improvement to the training speed of neural network. Open source GPU computing framework is also constantly innovative, promoting in-depth learning in all aspects of the application, Caffe is one of them. Because of its simple and easy to use and powerful performance, Caffe framework has been widely recognized. The Caffe framework is used to identify the seal types, and the two models have achieved excellent experimental results, which provides a new reference for the automatic recognition of seals.
【作者單位】： 四川大學(xué)計(jì)算機(jī)學(xué)院;
【基金】：國家自然科學(xué)基金資助項(xiàng)目(61303015) 四川省科技計(jì)劃項(xiàng)目(No.2014GZ0005-5)
【分類號】：TP391.41

【參考文獻(xiàn)】

相關(guān)碩士學(xué)位論文 前1條

1 尚利峰;脈沖耦合神經(jīng)網(wǎng)絡(luò)在圖像處理中的應(yīng)用[D];電子科技大學(xué);2007年

【共引文獻(xiàn)】

相關(guān)碩士學(xué)位論文 前2條

1 陳龍斌;基于脈沖耦合神經(jīng)網(wǎng)絡(luò)的圖像分割與圖像融合研究[D];云南大學(xué);2015年

2 胡芳;脈沖耦合神經(jīng)網(wǎng)絡(luò)在圖像分割和人臉檢測中的應(yīng)用研究[D];云南大學(xué);2011年

，

本文編號：2500433