Creating hdf5 files from .obj files

[Komti13]

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

[czq142857]

This repo does not contain code or instructions to create hdf5 files from obj files. Did you use the code from somewhere else?

Anyway, please make sure your input data (the binary voxels) are not emply. If they are not empty, the output hdf5 files should not be empty.

[Komti13]

i see so the only input type i can use is binary voxels and if i want to use obj files (meshes) i should convert them to binary voxels?

[czq142857]

Yes. You can read other issues for how to convert obj into binvox.

[Komti13]

okay thanks a lot for your help

[Komti13]

Excuse me i did the conversion to .binvox files but there this error in code of .mat files extension which i dont know what it stands for

can you please tell me what are the .mat files that i should get to avoid this error
and doing me a big favour can you inform me what i need as input (and where to put input files) to get all the data preparation files executed and the hdf5 files generated successfully. because in the code i found many different data paths and i am confused between them
voxel_input = /local-scratch/zhiqinc/shapenet_hsp/modelBlockedVoxels256/
image_input = /home/zhiqinc/zhiqinc/shapenet_atlas/ShapeNetRendering/
voxel_input_dir = "/home/zhiqinc/zhiqinc/shapenet_hsp/HSPvoxels256/"
img_input_dir = "/home/zhiqinc/zhiqinc/shapenet_hsp/HSPvoxels/"

[Komti13]

I tried to change the code and open the binvox file like following

and i get this error

ps: i tried another solution which is to convert .binvox to .mat file with this script xingyuansun/pix3d#19 (comment)
but the same error mentioned above

[czq142857]

You only need to assign the voxel grid to voxel_model_256. So here's what you should do.

Replace the following

		#get voxel models
		try:
			voxel_model_mat = loadmat(name_list[idx][1])
		except:
			print("error in loading")
			exit(-1)
		
		
		
		voxel_model_b = voxel_model_mat['b'][:].astype(np.int32)
		voxel_model_bi = voxel_model_mat['bi'][:].astype(np.int32)-1
		voxel_model_256 = np.zeros([256,256,256],np.uint8)
		for i in range(16):
			for j in range(16):
				for k in range(16):
					voxel_model_256[i*16:i*16+16,j*16:j*16+16,k*16:k*16+16] = voxel_model_b[voxel_model_bi[i,j,k]]
		#add flip&transpose to convert coord from shapenet_v1 to shapenet_v2
		voxel_model_256 = np.flip(np.transpose(voxel_model_256, (2,1,0)),2)

with

		voxel_model_file = open(name_list[idx][1], 'rb')
		voxel_model_256 = binvox_rw.read_as_3d_array(voxel_model_file).data.astype(np.uint8)
		#you need to make absolutely sure that the top direction of your shape is j-positive direction in the voxels,
		#otherwise the z-carving code will not work properly. (z-carving ≈ floodfill to make the voxels solid inside)
		#you can adjust the direction of your voxels via np.flip and np.transpose.

[Komti13]

Thank you for your response it really helps, but unfortunately the error this time that the voxel is not of length 256. so it raise an error of index out of bounds

the object voxel_model_256 has a length of 64.

[czq142857]

So just prepare 256^3 voxels.

[Komti13]

Okay thank you for helping