still issues about generate data .

[Myzhencai]

hi, @MathGaron as you mentioned that I need to have a look of the Opengl and I do reinstall everything in a new computer but i still get the same issues .here in pic1 I used to test the opengl it is fine:

and then i run it with the skull ply files then i get this : with no error no warning and what I think it can not be the json file issue as you can see in the picture .could it be the issue of plyparser or any thing you do not mention in the project?BTW,could you share the workflow of install dependences or just share the generated data .That can be of great help with my issue.

and in the plyparser.py this part never used in the code why should we import PlyElement?
wish for your feedback,I only have just 3 days for the project .
btw,can the generate black picture be the training data as it cotains the pose information?
"""
@staticmethod
def save_points(points, path):
vertex = np.zeros(points.shape[0], dtype=([('x', 'f4'), ('y', 'f4'), ('z', 'f4')]))
vertex.fill(255)
vertex['x'] = points[:, 0]
vertex['y'] = points[:, 1]
vertex['z'] = points[:, 2]
el = PlyElement.describe(vertex, 'vertex')
PlyData([el], text=ascii).write(path)
"""

[Myzhencai]

When i add the print(rgbB), I got this things

[MathGaron]

you can view the images with matplotlib or cv2.imshow, it is easier than looking at the print.

cv2.imshow("test", rgb[:, :, ::-1])
cv2.waitKey()
# or
plt.imshow(rgb)
plt.show()
plt.imshow(depth)
plt.show()

I am sorry but I can hardly send you an already generated dataset right now, I am working on another version and approaching a deadline.

That said, it seems that most of your setup works, but it seems like the 3D model is not loaded. Your json files seem also fine.

You could try this camera.json file instead :

You can verify that in modelrenderer.py, setup buffer, the vertex information are properly loaded.

[Myzhencai]

@MathGaron ,thanks for help.I will try something more with your suggestions.Have a nice day.

[Myzhencai]

@MathGaron here is the every line start from . i think it is not the issue about your code ,i can be the gpu issue as there are many buffers for gpu ,and my gpu is like this .can you check something unnormal in the workflow.Maybe i need to change another gpu to have a try.BTW,i do changed the camera.json file and it is the same issue and the code to absorbed the change of the json as the widows changed to the size of the camera parameters.looking forward your feedback.thanks : -).
for model in MODELS: vpRender = ModelRenderer(model["model_path"], SHADER_PATH, dataset.camera, window, window_size) vpRender.load_ambiant_occlusion_map(model["ambiant_occlusion_model"])

[MathGaron]

Ok the issue seems to be with the normal buffer.
Actually, you are not supposed to receive any NAN values in there, that would explain the weird triangles that you get as output I guess.

So first, make sure that you have the same model files (.ply) as the one on our website, and/or make sure that the 4 objects give you the same problem. If you still have the problem, then there is something wrong in the code.

As I told you earlier, I think I only tested with python3, so if you are using python2, it is possible that you get weird numerical bugs like this one.

If you still want to use python2 ( But I strongly suggest using python3!) the bug is potentially related to the line 50 in modelrenderer.ply. At this point, I normalize the normal vectors, and for some reason, the vector norm could be 0 which generate a division by zero error.

[Myzhencai]

hi, @MathGaron , I use python3.4 and all the models have the same issue (i have checked the ply files in windows it is wonderful just as the paper and ppt)and as I install almost every dependences i get the mention that it is link to .....navida...,so i guess it must be gpu issue .cause it can not compute out the right number.any way, i will try in a workstation with gpu quadrom2000m . i will share the result with you tomorrow. have a nice day .it is 22:36 in china

[MathGaron]

Hey,
I don't think it is a gpu related bug. I would focus on understanding why do the normals contain nan values.
So the bug seems to be related to the 3D model loading, and not with the rendering. Make sure that the function loads the normals correctly, and if they do, make sure that the normalization is fine and contains no 0 norms.

[Myzhencai]

hi, @MathGaron ,according to your guide I just tested the "he bug is potentially related to the line 50 in modelrenderer.ply."here is the result it seems it is the overflow of this line make all issues.I try to change the type to np.float64 but failed with missing or with can not find nx, ny,nz .Do you think this is a issue with code or i just need to change the platform (gpu) for generate data?

[MathGaron]

I am not sure what you are printing in your screenshots, but it seems like it is the normals after calling plyparser. I am also not sure why there are more than 3 values on axis 1. That said, there are huge values that should not be there (xe-38). I will try to get some time today to reproduce this and update you later.

[Myzhencai]

hi,the main of all the thing i did is that i first use the test.ply (made by myself ) to test the normals after calling plyparser then i print the normal after this code:
'''normals = normals / np.linalg.norm(normals, axis=1)[:, np.newaxis]'''
and no matter which ply file i change it still show the error like the image, but when i changed the code in plyparser in this code :
"""
def get_vertex_normals(self):
for element in self.data.elements:
if element.name == "vertex":
try:
return PlyParser.recarray_to_array(element.data[["nx", "ny", "nz"]], np.float32)(i changed it to np.float64)
"""
as you can see in the code i marked with red ,when i do not change it it show 3 number with warning ,when i changed it ,then it show 2 number without warning but missing one.

[Myzhencai]

and sorry for that , i can not run in the workstation as i mentioned cause my classmates just make the system crashed by installing the cuda.oh my god!!

[MathGaron]

Can you send your custom .ply, there is something wrong with it. The normals should be a Nx3 array with information about every vertex normals. Currently, I can't reproduce your bug.

[Myzhencai]

OK，i will share to
you tomorrow will， acturely I just modified the geometry.ply of dragging. Change d the nubumer of element to 5 and the nx ny nz to 1 2 3 ...

[Myzhencai]

And I just set print normal[0:1,I do not remember the data of this setting, but It just make the printing is just the three normal value]

[Myzhencai]

ply format ascii 1.0 comment VCGLIB generated element vertex 5 property float x property float y property float z property float nx property float ny property float nz property uchar red property uchar green property uchar blue property uchar alpha property float quality element face 2 property list uchar int vertex_indices end_header -0.0758073 0.0147489 -0.00880594 1 2 3 191 191 191 255 0.19343 -0.0757646 0.0144919 -0.00830121 4 5 6 175 175 175 255 0.177557 -0.0761136 0.0143482 -0.00857608 7 8 9 192 192 192 255 0.193868 -0.0761937 0.015855 -0.0102308 10 11 12 184 184 184 255 0.185957 -0.0760448 0.0155081 -0.00982756 13 14 15 192 192 192 255 0.194475 3 39499 39298 39485 3 37863 39485 39298

just little change of the ao.ply for 3dmodels of dragon.