Ask the Basel Face Model 2019 version first

[anleleaaa]

Hello, I would like to ask how to generate random Face using Basel Face Model 2019 version (that version is a.h5 file, I will not use),
I would be very grateful if you can help me，That version is a full human head

[BernhardEgger]

To generate a random face please run the software as described in the readme file.
You can then press the lower "random" button to generate a random face - with the upper random button you can also only generate random shape, random color or random expression.

To use the BFM 2019 you just select the corresponding model file when starting the tool instead of using the face only model file.

[anleleaaa]

要生成随机面孔，请按照自述文件中的说明运行该软件。 然后，您可以按下方的“随机”按钮生成随机面孔 - 使用上方的随机按钮，您也只能生成随机形状、随机颜色或随机表情。

要使用 BFM 2019，您只需在启动工具时选择相应的模型文件，而不是使用仅面部模型文件。

Can we add a friend? My email address is 1497840318@qq.com. I am a user from China, and my current major is 1. My research direction is 3D face reconstruction

[anleleaaa]

要生成随机面孔，请按照自述文件中的说明运行该软件。 然后，您可以按下方的“随机”按钮生成随机面孔 - 使用上方的随机按钮，您也只能生成随机形状、随机颜色或随机表情。

要使用 BFM 2019，您只需在启动工具时选择相应的模型文件，而不是使用仅面部模型文件。
I used the 2019 version of 3dmm before, and only successfully generated 3D point clouds in matlab, but failed to generate the real grid of human face. Few people in my university study this direction, so I really hope to get your help. At present, what I want to do is to randomize parameters to generate random heads (2019 version is full heads) and then do self-monitoring training. If you can help me, I can share my code with you (thank you very much).

[BernhardEgger]

You can export a mesh in PLY format from the tool - I hope that helps

[anleleaaa]

Hello, (I don't want to ask how to use this software) I mainly want to know how to generate random face steps, my purpose is to generate different faces according to the random number vector (199 dimension) and then use the formula, and then project this face to the 2 dimension into a picture, Then (image feature point extraction) is sent into the neural network to predict the random vector (199 dimensions) in the reconstruction of the face, this is the purpose of my current research, is also my postgraduate thesis idea, so that the self-supervised training of picture face reconstruction of the neural network, this is what I need (I do not want to ask you about the specific use of this software, Because I see your program screenshot effect is good, so I ask you to know) the problem I now encounter is that in matlab I initialize random number generated basically are almost the same face, if the random vector multiplied by several times can not get the generated object is not a face at all, I want to ask is your random vector is how to set? Here is my code and the running image If I set the mean of that shape vector to 1 it's not going to look good

…

------------------ 原始邮件 ------------------ 发件人: "unibas-gravis/basel-face-model-viewer" ***@***.***>; 发送时间: 2023年1月12日(星期四) 晚上9:11 ***@***.***>; ***@***.******@***.***>; 主题: Re: [unibas-gravis/basel-face-model-viewer] Ask the Basel Face Model 2019 version first (Issue #24) You can export a mesh in PLY format from the tool - I hope that helps — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[BernhardEgger]

When you want to sample from the Basel Face Model according to the prior you can use Gaussian distributed random values (the reasonable range of resulting values is between -3 and 3)
Your code and images did not come through

[anleleaaa]

Please wait a moment, I will send the source file to you, let you have a look, the compressed package is about 300 M

[BernhardEgger]

Sorry but this is not the kind of support we can offer here - we do not have capacity to look at your complete code package We do provide a reference implementation and you can find it and its documentation here: https://github.com/unibas-gravis/scalismo-faces/blob/master/src/main/scala/scalismo/faces/momo/MoMo.scala e.g. in line 340 for one of the directions. The necessary equations are also in the BFM paper and in the original 3DMM paper You might miss to multiply with the eigenvalues/variances from the hdf5 file?

…

-- 0 1 0 0 0 1 1 1 1 Am Fr., 13. Jan. 2023 um 13:27 Uhr schrieb anleleaaa < ***@***.***>:

Please wait a moment, I will send the source file to you, let you have a look, the compressed package is about 300 M — Reply to this email directly, view it on GitHub <#24 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFZK3YSPPZ5BD6LG3S2TXPTWSFC3DANCNFSM6AAAAAATYGEWKI> . You are receiving this because you commented.Message ID: ***@***.***>

[anleleaaa]

如果您没有经历看那个代码可以看下这个图片，但是我的代码也真的不多就一页 这是让均值向量乘不同的倍数

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人: "unibas-gravis/basel-face-model-viewer" ***@***.***&gt;; 发送时间:&nbsp;2023年1月13日(星期五) 晚上8:38 ***@***.***&gt;; ***@***.******@***.***&gt;; 主题:&nbsp;Re: [unibas-gravis/basel-face-model-viewer] Ask the Basel Face Model 2019 version first (Issue #24) Sorry but this is not the kind of support we can offer here - we do not have capacity to look at your complete code package We do provide a reference implementation and you can find it and its documentation here: https://github.com/unibas-gravis/scalismo-faces/blob/master/src/main/scala/scalismo/faces/momo/MoMo.scala e.g. in line 340 for one of the directions. The necessary equations are also in the BFM paper and in the original 3DMM paper You might miss to multiply with the eigenvalues/variances from the hdf5 file?

-- 0 1 0 0 0 1 1 1 1 Am Fr., 13. Jan. 2023 um 13:27 Uhr schrieb anleleaaa < ***@***.***&gt;: &gt; Please wait a moment, I will send the source file to you, let you have a &gt; look, the compressed package is about 300 M &gt; &gt; — &gt; Reply to this email directly, view it on GitHub &gt; <#24 (comment)&gt;, &gt; or unsubscribe &gt; <https://github.com/notifications/unsubscribe-auth/AFZK3YSPPZ5BD6LG3S2TXPTWSFC3DANCNFSM6AAAAAATYGEWKI&gt; &gt; . &gt; You are receiving this because you commented.Message ID: &gt; ***@***.***&gt; &gt; — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***&gt;

[BernhardEgger]

I removed your message as it contained a link redistributing the Basel Face Model which is not allowed according to the license. Please remove the provided zip-file from the server you uploaded it!

[BernhardEgger]

I removed your message again as it contained a link redistributing the Basel Face Model

[anleleaaa]

我再次删除了你的消息，因为它包含一个重新分发巴塞尔面部模型的链接

稍等，我把图片的发给你

[anleleaaa]

要生成随机面孔，请按照自述文件中的说明运行该软件。 然后，您可以按下方的“随机”按钮生成随机面孔 - 使用上方的随机按钮，您也只能生成随机形状、随机颜色或随机表情。

要使用 BFM 2019，您只需在启动工具时选择相应的模型文件，而不是使用仅面部模型文件。

[BernhardEgger]

The only difference I see is that you multiply the shape mean with a different constant? I guess this is to handle the difference between milimeter (bfm2019) and micrometer of bfm09? In that case tha factor should for both examples be 1000 and also for the shape pca you should multiply by 1000

…

-- 0 1 0 0 0 1 1 1 1 Am Fr., 13. Jan. 2023 um 13:54 Uhr schrieb anleleaaa < ***@***.***>:

要生成随机面孔，请按照自述文件中的说明运行该软件。 然后，您可以按下方的“随机”按钮生成随机面孔 - 使用上方的随机按钮，您也只能生成随机形状、随机颜色或随机表情。 要使用 BFM 2019，您只需在启动工具时选择相应的模型文件，而不是使用仅面部模型文件。 [image: 捕获 2PNG] <https://user-images.githubusercontent.com/44681546/212324694-5c6ac04f-3439-4aa8-b0b2-5d9046fc7770.PNG> [image: 捕获] <https://user-images.githubusercontent.com/44681546/212324718-1502ce4c-c80b-453c-8e45-f14a548655f6.PNG> — Reply to this email directly, view it on GitHub <#24 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFZK3YRYB4GIPPN4QEEPZALWSFF65ANCNFSM6AAAAAATYGEWKI> . You are receiving this because you commented.Message ID: ***@***.***>

[anleleaaa]

Very uncomfortable, the generated shape is distorted. If the shape_pca_vair variable is not 1000 times, it would be a normal face, but it would be a face every time I run it (I expect it to be a different face with different random numbers every time I run it).

[anleleaaa]

我看到区别您将形状均值乘不同的的常数？想想这这（BFM2019）和bfm09和和和是是乘乘乘乘乘乘乘乘的的的想想，形状 pca 也应乘以 1000
……
-- 0 1 0 0 0 1 1 1 1 Am Fr., 13. Jan. 2023 um 13:54 Uhr schrieb anleleaaa < @.>:
生成面孔，请按照文件中的运行软件。然后然后，您。然后然后然后然后。。可以可以按按按按下方下方的的的的的的按按按按生成生成随机面孔面孔面孔使用使用使用上方。 要使用BFM 2019，您只需要在启动工具时选择相应的模型文件，而不是使用仅部分模型文件。 [image: 捕获2PNG] < https ://user-images.github429123/4614-26 5c6ac04f-3439-4aa8-b0b2-5d9046fc7770.PNG > [image: 捕获] < https://user-images.githubusercontent.com/44681546/212324718-1502ce4c-c80b-453c-8e45-f14a548655f6.toPNG > - 回复直接发送电子邮件，在 GitHub < #24 (comment) > 上查看，或取消订阅 < https://github.com/notifications/unsubscribe-auth/AFZK3YRYB4GIPPN4QEEPZALWSFF65ANCNFSM6AAAAAATYGEWKI >。您收到此消息是因为您发表了评论。消息 ID：@.>

Are you interested in doing this small project? If I issue a paper, I must cite your project, O(∩_∩)O

[anleleaaa]

我看到区别您将形状均值乘不同的的常数？想想这这（BFM2019）和bfm09和和和是是乘乘乘乘乘乘乘乘的的的想想，形状 pca 也应乘以 1000
……
-- 0 1 0 0 0 1 1 1 1 Am Fr., 13. Jan. 2023 um 13:54 Uhr schrieb anleleaaa < @.>:
生成面孔，请按照文件中的运行软件。然后然后，您。然后然后然后然后。。可以可以按按按按下方下方的的的的的的按按按按生成生成随机面孔面孔面孔使用使用使用上方。 要使用BFM 2019，您只需要在启动工具时选择相应的模型文件，而不是使用仅部分模型文件。 [image: 捕获2PNG] < https ://user-images.github429123/4614-26 5c6ac04f-3439-4aa8-b0b2-5d9046fc7770.PNG > [image: 捕获] < https://user-images.githubusercontent.com/44681546/212324718-1502ce4c-c80b-453c-8e45-f14a548655f6.toPNG > - 回复直接发送电子邮件，在 GitHub < #24 (comment) > 上查看，或取消订阅 < https://github.com/notifications/unsubscribe-auth/AFZK3YRYB4GIPPN4QEEPZALWSFF65ANCNFSM6AAAAAATYGEWKI >。您收到此消息是因为您发表了评论。消息 ID：@.>

Are you interested in doing this small project? If I issue a paper, I must cite your project, O(∩_∩)O

If you are interested in it, I will give you the separate code. You only need to modify the path of the file, which is just a script file. I think the complete head in 2019 version is better than the face in the previous version

[anleleaaa]

matlab&nbsp; code Thank you very much for your guidance Teacher, you can try it yourself. Without changing the random number, the result looks almost the same every time. &nbsp; That script, if you use it, requires the 2019 version of the.h5 file to be placed under the appropriate directory, &nbsp; Instead of using the cells variable (which is the vertex index matrix for each cell of the grid), you can also view the output code as follows by drawing the point cloud method &nbsp; &nbsp; &nbsp; colormap(color); &nbsp; cdata = 1:size(vertex_matrix,1); &nbsp; scatter3(vertex_matrix(:,1),vertex_matrix(:,2),vertex_matrix(:,3),[],cdata); &nbsp; colorbar;

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人: "unibas-gravis/basel-face-model-viewer" ***@***.***&gt;; 发送时间:&nbsp;2023年1月13日(星期五) 晚上9:03 ***@***.***&gt;; ***@***.***&gt;;"State ***@***.***&gt;; 主题:&nbsp;Re: [unibas-gravis/basel-face-model-viewer] Ask the Basel Face Model 2019 version first (Issue #24) The only difference I see is that you multiply the shape mean with a different constant? I guess this is to handle the difference between milimeter (bfm2019) and micrometer of bfm09? In that case tha factor should for both examples be 1000 and also for the shape pca you should multiply by 1000

-- 0 1 0 0 0 1 1 1 1 Am Fr., 13. Jan. 2023 um 13:54 Uhr schrieb anleleaaa < ***@***.***&gt;: &gt; 要生成随机面孔，请按照自述文件中的说明运行该软件。 然后，您可以按下方的“随机”按钮生成随机面孔 - &gt; 使用上方的随机按钮，您也只能生成随机形状、随机颜色或随机表情。 &gt; &gt; 要使用 BFM 2019，您只需在启动工具时选择相应的模型文件，而不是使用仅面部模型文件。 &gt; &gt; [image: 捕获 2PNG] &gt; <https://user-images.githubusercontent.com/44681546/212324694-5c6ac04f-3439-4aa8-b0b2-5d9046fc7770.PNG&gt; &gt; [image: 捕获] &gt; <https://user-images.githubusercontent.com/44681546/212324718-1502ce4c-c80b-453c-8e45-f14a548655f6.PNG&gt; &gt; &gt; — &gt; Reply to this email directly, view it on GitHub &gt; <#24 (comment)&gt;, &gt; or unsubscribe &gt; <https://github.com/notifications/unsubscribe-auth/AFZK3YRYB4GIPPN4QEEPZALWSFF65ANCNFSM6AAAAAATYGEWKI&gt; &gt; . &gt; You are receiving this because you commented.Message ID: &gt; ***@***.***&gt; &gt; — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you modified the open/close state.Message ID: ***@***.***&gt;