Comments (9)
Hi @VERMANs 首先你的ttn的风格数据似乎是有点少,论文里是1w张生成+真实的风格图
你的ttn训练的时候的真实图片是经过align和crop的人脸吗?你的ttn做inference的时候如果只是针对人脸部分会好一些吗?如果这些都没有问题的话,尝试着加大batch(分布式训练,增加节点数),增加迭代次数,调整一下损失函数的比重,也许会有不错的结果
祝你好运!
from dct-net.pytorch.
Hi @LeslieZhoa 非常感谢你的回复,不好意思,我上面表达可能有误,我在ttn训练时的风格数据是生成的2w风格图(没有用到真实的风格图,这操作对的吗),
ttn训练的真实图片是经过align和crop的人脸,如下所示
在ttn只对人脸效果也并不理想(且是训练集的人脸),效果如下
对于你的建议,我会尝试,后续有结果,我会及时跟你反馈!
from dct-net.pytorch.
@VERMANs ttn的风格数据是要再经过人工筛选的,不知道你是不是有操作这一步
from dct-net.pytorch.
@LeslieZhoa 是的,我因为偷懒没做这步,可是这步是需要从2w张人工挑选5000-1w张吗,我想知道这里面好坏的标准大概是什么?
from dct-net.pytorch.
@LeslieZhoa 是的,我因为偷懒没做这步,可是这步是需要从2w张人工挑选5000-1w张吗,我想知道这里面好坏的标准大概是什么?
@VERMANs 纯主观判断,就是你觉得该图片生成的风格可以接受就ok
from dct-net.pytorch.
@LeslieZhoa 好的,非常感谢,我在尝试,有结果第一时间与你反馈
from dct-net.pytorch.
你好 @LeslieZhoa
下面是结果,(图片依次为:测试图片1,测试结果1,测试图片2,测试结果2,风格图样例1,风格图样例2)
请问这个效果是正确的吗?关于测试结果2,他衣服的颜色风格是否来源于样式图中人像以外的白色背景呢?
期待你的回复,愿你万事如意
from dct-net.pytorch.
@VERMANs 我感觉是你的风格图片全是偏白色背景,导致模型无法学习风格化的背景,所以脸部周围会有些artifacts,衣服颜色会变成白色。一点小建议,如果想继续沿用这种风格的话,随意找一些正常背景图片,加上opengl的lut处理,再和分割后的风格人脸进行拼接,这样训练也许会对背景生成有帮助。
祝你好运!
from dct-net.pytorch.
@VERMANs Is image pairing required during training?
from dct-net.pytorch.
Related Issues (15)
- Great job! HOT 1
- Does the ccn stage need to be trained for so long? HOT 1
- Why doesn't the pairing of data generated by get_tcc_input.py look like?
- error: name 'Engine' is not defined HOT 1
- 一个笔误在model/Pix2PixModule/module.py中的color_shift HOT 1
- 背景问题
- 请问CCN模型和TTN模型在使用上有什么区别?为什么要训练两个模型?
- Inference does not work correctly HOT 4
- Training data size HOT 3
- Training error HOT 1
- 可以跑训练了,取消issues HOT 1
- 关于ccn的blend和eometry expansion module
- 非常好的开源代码
- The inferred effect is black?
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from dct-net.pytorch.