huailiang / face-nn Goto Github PK

View Code? Open in Web Editor NEW

473.0 17.0 143.0 129.44 MB

游戏捏脸，基于神经风格迁移框架生成逼真人脸

Home Page: https://huailiang.github.io/blog/2019/face/

License: MIT License

ShaderLab 3.15% C# 32.05% HLSL 32.18% Shell 0.01% Mask 3.57% Python 24.14% C 0.17% C++ 1.69% Cuda 3.05%

unity game-engine face-recognition neteasecloud neural-network cnn imitation-learning

face-nn's People

Contributors

Stargazers

Watchers

Forkers

bmjoy hengle sunmars mengtest cfengfeng arixlin bluesea maple-yang eleven8023 superowner eeii yaoqingyuan ggdebbs flycodes arikes yudixiaok whztt07 wenboqiu oolijianoo flyarong leelnfei lingkeyang xiyuhao teejs2012 xiongnc abstractor97 linrayx doubtfisher frankiegu nibaole str818 cyecp michaelqiu1992 ly774508966 qiezibao vertigooooo bestjex xieweibo buffernihility seem-sky hafewa apathetically kuustudio alec-ray liuguoyou lgqfhwy peterzs gtechx mealiu lzhbrian scholltan zhyj3038 learnmorecode linzai1992 i3nli 1165048017 zcvdf jo-dean kobemy31 cshopping bvirgo 351zyf mowenli arthur-qi waylandgod lryain klj35 foryun1009 hozhis itana lith0613 mathbee lastexile-ch 178435865 swordlegend dongyaolin tt-52101 chengkunxf learningpro gofirestar d34d633f zhipeng515 elongame gaolf hustllz 18724799167 waella blessing2012 wikieden shineway14 111304037 meadowair declanctz simpletuo dongdong168 kerwinzxc haoduoyu1203 azusa306 xiaoli9 roy2an

face-nn's Issues

Feature Extractor

我看到代码里面有“train_extractor”这个过程，但是在论文里面我没有找到相关介绍

你能说明一下这个网络结构吗？不胜感激！

CFUtilPoolLib这个dll是干嘛用的啊，为什么XRoleParts组件继承MonoBehavior却在这个dll中

CFUtilPoolLib.dll主要处理的是哪个过程？

UnauthorizedAccessException in unity when generating dataset

Help, i don't how to solve this problem in unity. Baidu tells me to change the code. But i'm not sure.

是否可以基于3D头像来进行捏脸？

通过3D头像数据，数据本身包含深度信息，通过捏脸来实现不同的风格变换。
我这个头像就是3D扫描出来的，然后戴上虚拟的眼镜。我想在这个3D扫描的基础上在捏脸，改变太写实的人脸造型，这个是否会更好？

请问下这个../dat/79999_iter.pth是怎么来的呢
File "evaluate.py", line 130, in <module> img_1 = faceparsing_ndarray(image, '../dat/79999_iter.pth', cuda) File "evaluate.py", line 91, in faceparsing_ndarray build_net(cp, cuda) File "evaluate.py", line 81, in build_net _net_, _to_tensor_ = _build_net(cp, cuda) File "evaluate.py", line 61, in _build_net net.load_state_dict(torch.load(cp, map_location="cpu")) File "/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/torch/serialization.py", line 382, in load f = open(f, 'rb') FileNotFoundError: [Errno 2] No such file or directory: '../dat/79999_iter.pth'

兄弟你用的不是helen数据集啊，你语义分割用的是CelebAMask-HQ数据集啊....

想问下 phase=evaluate这个阶段的问题

想问下evaluate 这个阶段的操作没有开源嘛？

error

Hi，when I do inference, it has errors like this:size mismatch for 0.0.weight: copying a param with shape torch.Size([95, 512, 4, 4]) from checkpoint, the shape in current model is torch.Size([99, 512, 4, 4]). If I change params_cnt from 99 to 95, it will has a new error:index 96 is out of bounds for dimension 1 with size 95, because of self.argmax_params(t_params.data, 96, 3). What should I do? Changing argmax_param or could you please share model which is 99 dimension. Looking forward to your reply!@huailiang