【导读】Github上有许多成熟的TensorFlow代码和模型,可以直接用于科研和工程中。本文会介绍其中一些与自动问答、机器翻译、风格转换等相关的精致且实用的项目。
Seq2Seq Chatbot
200行代码实现的Twitter/Cornell-Movie聊天机器人
项目地址:
https://github.com/tensorlayer/seq2seq-chatbot
相关资料:
Practical-Seq2Seq:
http://suriyadeepan.github.io/2016-12-31-practical-seq2seq/
The Unreasonable Effectiveness of Recurrent Neural Networks:
http://karpathy.github.io/2015/05/21/rnn-effectiveness/
Understanding LSTM Networks:
http://colah.github.io/posts/2015-08-Understanding-LSTMs/
运行样例:
Query > happy birthday have a nice day
> thank you so much
> thank babe
> thank bro
> thanks so much
> thank babe i appreciate it
Query > donald trump won last nights presidential
debate according to snap online polls
> i dont know what the fuck is that
> i think he was a racist
> he is not a racist
> he is a liar
> trump needs to be president
Domain Transfer Network (DTN)
跨域图像生成,例如输入真人照片,生成卡通头像。
项目地址:
https://github.com/yunjey/domain-transfer-network
相关论文:
《Unsupervised Cross-Domain Image Generation》:
https://arxiv.org/abs/1611.02200
neural-style-tf
利用卷积网络变换图像的风格,将普通图像变为素描、艺术画等。
项目地址:
https://github.com/cysmith/neural-style-tf
参考文章:
《Image Style Transfer Using Convolutional Neural Networks》:
http://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Gatys_Image_Style_Transfer_CVPR_2016_paper.pdf
《Artistic style transfer for videos》:
https://arxiv.org/abs/1604.08610
《Preserving Color in Neural Artistic Style Transfer》:
https://arxiv.org/abs/1606.05897
darkflow
YOLO目标检测,可以直接导入Darknet的模型。
项目地址:
https://github.com/thtrieu/darkflow
相关论文:
YOLOv1:
https://arxiv.org/abs/1506.02640
YOLOv2:
https://arxiv.org/abs/1612.08242
transformer
论文Attention is All you Need的实现,端到端机器翻译,可以在自己电脑上实现谷歌翻译的功能。
项目地址:
https://github.com/thtrieu/darkflow
相关论文:
《Attention is All you Need》:
https://papers.nips.cc/paper/7181-attention-is-all-you-need.pdf
运行样例(输入德语,翻译得到英语):
source: Sie war eine jährige Frau namens Alex
expected: She was a yearold woman named Alex
got: She was a woman named yearold name
source: Und als ich das hörte war ich erleichtert
expected: Now when I heard this I was so relieved
got: And when I heard that I was an
-END-
专 · 知
人工智能领域26个主题知识资料全集获取与加入专知人工智能服务群: 欢迎微信扫一扫加入专知人工智能知识星球群,获取专业知识教程视频资料和与专家交流咨询!
请PC登录www.zhuanzhi.ai或者点击阅读原文,注册登录专知,获取更多AI知识资料!
请加专知小助手微信(扫一扫如下二维码添加),加入专知主题群(请备注主题类型:AI、NLP、CV、 KG等)交流~
请关注专知公众号,获取人工智能的专业知识!
点击“阅读原文”,使用专知