czq142857 / im-net Goto Github PK

May I ask if you could share with me the dataset for each category, which includes voxels, query points, point clouds and rendering picture?

Given, I have downloaded the pre-trained model and weights from the given link below :
https://github.com/czq142857/IM-NET#datasets-and-pre-trained-weights

If I want to run the test on a new image for generating mesh model, will the code work just by changing the below block with my own input_image_path and reading the image?
https://github.com/czq142857/IM-NET/blob/master/IMSVR/model.py#L489-L491

Do I need to do some more processing for the new input image?

While checking the link for data preparation,
https://github.com/czq142857/IM-NET#datasets-and-pre-trained-weights

If I have 3D mesh models, what steps should I follow to convert the dataset to point-value pairs?

Hi Zhiqin,

Thanks for the great work.
Can you provide the codes you use of the MSE, IoU, symmetric Chamfer distance (CD) for model evaluation?

Best regards,
czsc

I was checking the paper and playing with the code, and I was curious about the details regarding the GAN version vs the autoencoder version. The supplement material seems to show better interpolation results for GAN, but in the code I see that the GAN version (ZGAN class in modelz.py) is not even used for inference in the main, and instead it just used to fetch a random z and fed it to the AE. Any particular reason for this?

Also wights for the IM-GAN are not provided, am I correct?

Hi Zhiqin,

Thanks for the great work.

I follow your suggestions in other issues that i normalize each shape so that the diagonal of its bounding box equals to unit length.Then i use binvox.exe to voxelize the mesh to voxel. But i find the binvox.exe still rescale and translate the bounding box into a unit cube. And i found the example .binvox voxel in your original sampling points directory has a unit length diagonal .

if you use the binvox.exe to voxelize the mesh, could you please tell me the command that you use in terminal?
Or , if you use other voxelization tools or voxelization method, could you please share me or guild me how to voxelize the mesh ?

Best regards,
czsc

Greets, I like your work. In your paper, you sampled points on $16^3, 32^3, 64^3, 128^3$. However, in your code about point_sampling, you provided all mentioned resolutions except $128^3$, and the commands about training also ignored $128^3$. Could you share the point sampling code on $128^3$ or explain why the commands about training stage (autoencoding) do not use $128^3$? Thanks in advance.

Greets, I have another question: In your provide dataset (13 categories on ShapeNet), the training data are put together in one file all_vox256_img_train.hdf5. And I see the object's name in another file all_vox256_img_train.txt. Whether the object's name (in all_vox256_img_train.txt) corresponds to the data in all_vox256_img_train.hdf5 in order? For example, the data from the first line in all_vox256_img_train.txt corresponds to the first object from all_vox256_img_train.hdf5. Thanks in advance.

Hi Chen, thank you for sharing the code.
I am trying to train the network with my own data. But when I used 2_gather_256vox_16_32_64.py to generate training data from my 256^3 voxels, I got warnings like "64-- batch_size exceeded! and 32-- batch_size exceeded!" and "2_test_hdf5.py" generate something like this
3_p64_3
It seems something is wrong but I haven't found out the problem. Could you please give me some hints or have you met this before? I'm eager to know how to fix it.

Great Thanks!

Can you show the change of C code in LightField descriptor, i am so confused.

Hi @czq142857 kind regards, I want to train IMNET for other objects. Is it possible ? if yes how?
I have anohter query about data processing as you can mentioned that the data will be downloaded from HSP . So how we can prepare the input data to IMNET training for others objects.
as the output of IMNET point inside outside decision after you train the GAN model for shape generation. can we directly apply some techniques likes marching cubes to generate a shape from IMNET output.
Thanks for your assistance.

Hi, thank you for good code.

I want to know how can I generate my image using this code.
I think dataset "pixels" in "all_vox256_img_test.hdf5" is image. But I dont know how can I make dataset using my image.

Thank you

Hey, could you please tell me the difference between IM-Net and Occupancy Networks. Both look exactly the same to me.
It would be really helpful if you help me out with this question

Hello i was trying to create my own hdf5 dataset to use for ucsgnet. i want to generate the file from an .obj file extension but i keep getting empty hdf5 files please some help . i followed all the given instructions in the repository

Hi, how to get the same rendering results as this image? Thanks in advance!

When I tried to implement your code for IMGAN, I got an error (NameError for IMAE) as shown in the attached image. I have used the latest version of TensorFlow=2.0 and python=3.8, and Cuda=11.x, so if you could help me, please do so as soon as possible to fix this error because I'm a PhD student. I'm very interested in your work. Hopefully, I will get an answer very soon.

Hi,sincerely thank you for your work.
Did you know how can I make gif animation from a lot of .dae file. I'm using the great work in bionic design related research and I need to use it to produce the Interpolation effect.
I try many dae viewer but they can just view one model one time.
The origin author of the research say that he uses 3dmax in one issue and i guess it can be made by using 3dmax to make animation but i haven't taken it. I need more information about this.
thank you!

Hi, in issue #17 you mentioned the loss 0.07 is not right and it indicates all zero values for the points. what is the expected range of loss from the beginning to the end of training in each phase (approximately)? -trying to debug my pipeline-

@czq142857 : Thanks for the instructions. I followed the steps and trained my own model.
While using the trained model, I am getting all the dimensions of intermediate variables correctly.
But, I am getting an error in z2voxel code, after this step, model_float = self.z2voxel(model_z),
model_float is coming as all zeros values. What could be the possible error?

Thanks in advance.

Originally posted by @nitish11 in #8 (comment)

Dear Mr zhiqin,
Appreciate your great work and released codes!
Could you please help explain a little bit what variable "b" and "bi" represent in .mat file of HSP ? Thanks for your help!

Best,
Bowen