A Recipe for Training Neural Networks Andrej Karpathy blog 2019-04-27 09:37:05 This blog is copied from:https://karpathy.github.io/2019/04/25/recipe/ Some few weeks ago I posted a tweet on “the most common neural net mistakes”, listing a few
paper: 《Attention Augmented Convolutional Networks》 https://arxiv.org/pdf/1904.09925.pdf 这篇文章是google brain的,应该有分量。上来就说:卷积神经网络有一个重要的弱点就是 它仅仅操作于于一个领域,对于没有考虑到全局信息有损失。 (这就是全局和局部的辨证关系。) 注意力机制
FPN(feature pyramid networks)算法讲解 https://blog.csdn.net/u014380165/article/details/72890275/
QQ Group: 428014259 Tencent E-mail:403568338@qq.com http://blog.csdn.net/dgyuanshaofeng/article/details/89005078 [1] Exploring Randomly Wired Neural Networks for Image Recognition 2019 [paper]
[综述]Deep Compression/Acceleration深度压缩/加速/量化 Model-Compression-Papers Papers for neural network compression and acceleration. Partly based on link. Survey Recent Advances in Efficient Computation of Deep Convolutional Neural Networks, [arxiv '18] A S
导读 本文讨论了深层神经网络训练困难的原因以及如何使用Highway Networks去解决深层神经网络训练的困难,并且在pytorch上实现了Highway Networks。 一 、Highway Networks 与 Deep Networks 的关系 深层神经网络相比于浅层神经网络具有更好的效果,在很多方面都已经取得了很好
作者:O天涯海阁O 来源:CSDN 原文:https://blog.csdn.net/zhangjunhit/article/details/53261053 版权声明:本文为博主原创文章,转载请附上博文链接! OCR 资源汇总 字符区域检测:https://github.com/eragonruan/text-detection-ctpn 区域字符识别:https://github.com/meijieru/crnn.pyto