https://arxiv.org/pdf/1611.09326.pdf The One Hundred Layers Tiramisu: Fully Convolutional DenseNets for Semantic Segmentation https://harishnarayanan.org/writing/artistic-style-transfer/ http://www.ijiandao.com/2b/baijia/264053.html AI界最危险武器GPT-2使用指南:从Finetune到部署 风格转化视频
NEURAL NETWORKS FOR MODELING SOURCE CODE EDITS
图形分割标准数据级
数据 COCO:http://mscoco.org/
VOC pascal:http://host.robots.ox.ac.uk/pascal/VOC/
ADE20K:http://groups.csail.mit.edu/vision/datasets/ADE20K/
分布式部署
https://www.jianshu.com/p/bf17ac9e6357 https://www.jianshu.com/p/9c462bbb6628
相关链接
- 服务器端产品部署:
https://medium.com/@burgalon/deploying-your-keras-model-35648f9dc5fb
- 客户端产品部署:
https://becominghuman.ai/deploying-your-keras-model-using-keras-js-2e5a29589ad8
3.利用VGG网络实现图像分割:
4.FCN网络原论文:
https://arxiv.org/pdf/1411.4038.pdf
- Tiramisu网络论文:
https://arxiv.org/pdf/1611.09326.pdf
https://nbviewer.jupyter.org/url/files.fast.ai/part2/lesson14/tiramisu-keras.ipynb
- Tiramisu网络的代码实现:
http://files.fast.ai/part2/lesson14/
- 神经网络中37个常见问题:
https://blog.slavv.com/37-reasons-why-your-neural-network-is-not-working-4020854bd607
- 如何调试神经网络:
https://hackernoon.com/how-to-debug-neural-networks-manual-dc2a200f10f2
- 结合CNN网络和CRF方法的图像分割:
- 在你自己做出背景移除器之前,可以试试作者们的demo:
https://greenscreen-ai.boorgle.com/
最后,量子位想说,本教程适合学习,如果你只是想通过抠图讨女朋友欢心,Prisma团队后来推出的新应用Sticky是个不错的选择
https://github.com/soumith/ganhacks gan 技巧说明