Comments (11)
I know how to train custom data, just follow the readme steps.
from dinet.
Bit more difficult than that, loss convergence etc.. I tried to eyeball the results before moving onto each stage and the results did not match the work I had to put into collecting datasets etc, there are plenty of issues in this repo regarding similar issues which is why op has rather cleverly tried to avoid the headache/ learning curve.
from dinet.
Bit more difficult than that, loss convergence etc.. I tried to eyeball the results before moving onto each stage and the results did not match the work I had to put into collecting datasets etc, there are plenty of issues in this repo regarding similar issues which is why op has rather cleverly tried to avoid the headache/ learning curve.
I don't know what is your data looks like, but in my data, it seems work, athough this is some loss convergence issue, but it seems results not bad.
from dinet.
Bit more difficult than that, loss convergence etc.. I tried to eyeball the results before moving onto each stage and the results did not match the work I had to put into collecting datasets etc, there are plenty of issues in this repo regarding similar issues which is why op has rather cleverly tried to avoid the headache/ learning curve.
I don't know what is your data looks like, but in my data, it seems work, athough this is some loss convergence issue, but it seems results not bad.
Mine was a single person dataset so perhaps this was some over fitting etc, the dataset itself was high quality about 3 hours of front facing studio lighting, the results where a higher chin+cheek fidelity and less bounding box but jittery lips. Perhaps a more diverse dataset would have resolved this, I saw a lot of people asking about syncnet training and wasn't 100% aware of what that was
from dinet.
Bit more difficult than that, loss convergence etc.. I tried to eyeball the results before moving onto each stage and the results did not match the work I had to put into collecting datasets etc, there are plenty of issues in this repo regarding similar issues which is why op has rather cleverly tried to avoid the headache/ learning curve.比这更困难的是,损失收敛等。我试图在进入每个阶段之前观察结果,结果与我必须投入收集数据集等的工作不匹配,这个回购协议中有很多关于类似问题的问题这就是为什么 op 相当聪明地试图避免头痛/学习曲线。
I don't know what is your data looks like, but in my data, it seems work, athough this is some loss convergence issue, but it seems results not bad.我不知道你的数据是什么样的,但在我的数据中,它似乎有效,虽然这是一些损失收敛问题,但看起来结果还不错。
有尝试过中文的效果吗?
from dinet.
Bit more difficult than that, loss convergence etc.. I tried to eyeball the results before moving onto each stage and the results did not match the work I had to put into collecting datasets etc, there are plenty of issues in this repo regarding similar issues which is why op has rather cleverly tried to avoid the headache/ learning curve.比这更困难的是,损失收敛等。我试图在进入每个阶段之前观察结果,结果与我必须投入收集数据集等的工作不匹配,这个回购协议中有很多关于类似问题的问题这就是为什么 op 相当聪明地试图避免头痛/学习曲线。
I don't know what is your data looks like, but in my data, it seems work, athough this is some loss convergence issue, but it seems results not bad.我不知道你的数据是什么样的,但在我的数据中,它似乎有效,虽然这是一些损失收敛问题,但看起来结果还不错。
有尝试过中文的效果吗?
No but others have, but deepspeech wasn't trained on Chinese so the lip movements won't be fully accurate to Chinese audio
from dinet.
Bit more difficult than that, loss convergence etc.. I tried to eyeball the results before moving onto each stage and the results did not match the work I had to put into collecting datasets etc, there are plenty of issues in this repo regarding similar issues which is why op has rather cleverly tried to avoid the headache/ learning curve.比这更困难的是,损失收敛等。我试图在进入每个阶段之前观察结果,结果与我必须投入收集数据集等的工作不匹配,这个回购协议中有很多关于类似问题的问题这就是为什么 op 相当聪明地试图避免头痛/学习曲线。
I don't know what is your data looks like, but in my data, it seems work, athough this is some loss convergence issue, but it seems results not bad.我不知道你的数据是什么样的,但在我的数据中,它似乎有效,虽然这是一些损失收敛问题,但看起来结果还不错。
有尝试过中文的效果吗?
No but others have, but deepspeech wasn't trained on Chinese so the lip movements won't be fully accurate to Chinese audio
I tried to use the examples in the README for training and reasoning, but the lip tremors were severe, and the mouth still retained its original movements when there was no sound. So I think this project is not very good.
from dinet.
I'm not working for money, I'm working for fun. So you can try my project here it will be included training dinet full pipeline in several days..
https://github.com/primepake/better_wav2lip
from dinet.
I know how to train custom data, just follow the readme steps.
how to contact you? I would pay I need your help
from dinet.
Bit more difficult than that, loss convergence etc.. I tried to eyeball the results before moving onto each stage and the results did not match the work I had to put into collecting datasets etc, there are plenty of issues in this repo regarding similar issues which is why op has rather cleverly tried to avoid the headache/ learning curve.
I don't know what is your data looks like, but in my data, it seems work, athough this is some loss convergence issue, but it seems results not bad.
Mine was a single person dataset so perhaps this was some over fitting etc, the dataset itself was high quality about 3 hours of front facing studio lighting, the results where a higher chin+cheek fidelity and less bounding box but jittery lips. Perhaps a more diverse dataset would have resolved this, I saw a lot of people asking about syncnet training and wasn't 100% aware of what that was
how to contact you? I would pay I need your help
from dinet.
t full pipeline in several days..
#New Features: DINet full pipeline training
Very much looking forward to it!
from dinet.
Related Issues (20)
- Inference results in no audio sound, _add_audio.mp4 is not generated HOT 2
- 🤢 New repo LipSick 🤢🤮 HOT 1
- 有大佬成功训练出效果较好的 syncnet吗? HOT 8
- Model is changing Skin Tone
- 优化推理效果的讨论 HOT 38
- 求助,clip训练后,推理出来的视频会出现花屏帧?
- Who needs high-quality lip sync - contact me! HOT 14
- 关于visual quality的计算方式 HOT 2
- 可不可以不使用 openface? HOT 2
- 训练时出现奔溃帧,不知原因? HOT 4
- Tensorflow version 1.xx is PAIN (COLAB) HOT 1
- frame阶段训练loss出现倒刺现象 HOT 9
- I need a convergence chart for training HOT 2
- clip阶段sync loss的问题 HOT 2
- AI Dubbing API (with multi-speaker lipsync) HOT 1
- 关于训练过程中的验证
- num_workers 设为非 0 后报错
- frame 256的模型同步性很差 HOT 5
- Evaluation
- Is there a 128 Clip training model?
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 dinet.