deep-motion-editing

deep-motion-editing copied to clipboard

训练之后的效果存在问题

1

大神您好！非常感谢你们能够开源这么棒的工作，我最近也在做相关的工作，研究了一下你们的代码，尝试训练，运行的过程中遇到了问题，想请教一下。  上图是通过我们训练出来的结果得到的重定向数据，从左到右依次为gt，result，input。我们尝试去除了GAN_loss(gan mode设置为none)进行训练，训练用的dataset没有做过修改，但随着训练次数的增多，重定向result的末端关节会越来越接近source数据的状态，例如上图，Big Vegas的手脚和gt相比，明显倾向于Aj的input状态，请问这可能是什么原因导致的？ 此外，我还对loss的初始参数有点疑问，在option_parser.py中，lambda_ee的值默认为100，这与论文中的比例相差巨大，请问loss参数的这些默认值是否存在问题？ 希望大神能够抽空回答一下我的问题，感激不尽。

yutianhao-chingmu

I can't remove_fs because I'm getting NaN values after passing my poses into trainer.test(), my data is not coming from the examples that you are giving. Why is that the...

BaptisteLqt

About the 2d pose extraction in Openpose

2

Hi, thanks for sharing your great work! I'm curious about the style transfer in 2D space and want to learn more about it. Could you share the 2d pose extraction...

shinshiner

Joints order

2

Can you give the order of the joints and their names, because extracting it from the .bvh file is confusing with the end site. In what order do you give...

BptLqt

感谢您的工作！ 您在Train from scratch的Motion Style Transfer中提到 “The dataset consists of two parts: one is the taken from the motion style transfer dataset proposed by Xia et al. and the other...

Molicaliatimis

data pre-process

1

Hi guys, Thanks for your great work! I am a little bit confused about the data pre-process part of the retargeting project. When I test on a specific model, I...

RavenLeeANU

针对任意bvh模型的retargeting

8

hi，PeiZhuo，感谢你们非常棒的工作，我尝试运行训练和测试代码，都可以正常运行，当我尝试使用第三方bvh模型遇到以下问题： 1）如果要对第三方模型bvh重定向，必须重新训练？训练的话第三方模型的运动数据也要跟Mixamo一样吗，我的意思是要准备107个不同动作的bvh文件？第三方模型重定向的基本流程是怎样的？ 2）intra_structure和cross_structure是如何定义的？如果我仅仅对不同的humanoid 模型重定向是不是可以理解为intra_structure？ 3）输入模型先会做精简预处理，重定向后是简化后的模型，是不是还少了一步把inference出来的result结果恢复到简化前的模型上，这一步可以用skeleton unpooling 来实现吗？ 谢谢！盼回复！

cjlong1213

The bvh file generated by style transfer cannot be bound to the human body model

2

Hi, first of all thank you very much for your contribution, this is an excellent work! I am an animator and have been studying this project for some time. Recently,...

zanxiR

fix_foot_contact not fixing foot contact

1

Hi there! :) I just noticed that the feet of the resulting character 'go below zero' (left: input.bvh, right: result.bvh), even though the foot contact has been 'fixed' by the...

rainer7

train_problem

13

I modified `data_filename = "bfa.npz"` and `num_classes = 16` in style_transfer/config.py to train the bfa dataset, but the following error occurred: ``` mean and std saved at /home/ddd/deep-motion-editing/style_transfer/probe/../data/bfa_norms/train_style3d.npz data shape...

Hellodan-77