I recently wrote a Python binding for surface generation. During the test, I found tha

A few questions: Did you change the smoothing parameter to use

Two potential problem I encountered during using the code. about quickses HOT 9 CLOSED

nezix commented on August 18, 2024

Two potential problem I encountered during using the code.

from quickses.

Comments (9)

nezix commented on August 18, 2024

A few questions:

Did you change the smoothing parameter to use Laplacian smoothing ? This is the "l" parameter.
Did you change the voxel size, you can improve the quality of the mesh by setting it to something lower.
How do you visualize the results ? It looks like the normals are not used.

from quickses.

miemiemmmm commented on August 18, 2024

Thanks for your reply. I tried with a higher number of Laplacian smoothing, increased the slice size, and lowered the voxel size, the surface spike problem was reduced. However, the surface is still not water-tight.

The tool I used for visualization is Open3D. Basically, I just outputted the vertices and faces to the object file and the normal is from open3d's computation before rendering.

The surface you are viewing is the atoms within a cubic bounding box from part of a protein. As you might see atoms outside of the surface, they are atoms outside of the bounding box.

from quickses.

nezix commented on August 18, 2024

Slice size shouldn't matter as your surface seems rather small. If your GPU has enough memory to compute the whole surface in one run, you can ignore this setting.

I can increase the limit of the grid to see if that solves your issue.

from quickses.

nezix commented on August 18, 2024

Could you try this branch and see if that fixes your issue ?

from quickses.

miemiemmmm commented on August 18, 2024

There is still the problem in the fix-origin branch. I am using RTX4090 with 24GB memory. GPU hardware should not be the problem.

Could it be a Cuda version or nvcc problem? When I compile the program, there are some errors about missing functions. I have to add the following code to CudaSurf.cu to enable the compilation.

#include <thrust/sort.h>
#include <thrust/unique.h>
#include <thrust/count.h>

Error message: 
nvcc -c -o CudaSurf.o CudaSurf.cu -Icpdb/ -use_fast_math -O3
CudaSurf.cu(419): error: namespace "thrust" has no member "sort"
      thrust::sort(vertThrust, vertThrust + totalVerts, sort_float3());
              ^

CudaSurf.cu(421): error: namespace "thrust" has no member "unique"
      thrust::device_ptr<float3> last = thrust::unique(vertThrust, vertThrust + totalVerts, samefloat3());
                                                ^

CudaSurf.cu(567): error: namespace "thrust" has no member "sort"
      thrust::sort(D_beg, D_beg + N, cmp);
              ^

CudaSurf.cu(646): error: namespace "thrust" has no member "sort"
                  thrust::sort(fillThrust, fillThrust + sliceNbCellSES);
                          ^

CudaSurf.cu(648): error: namespace "thrust" has no member "count_if"
                  unsigned int notEmptyCells = thrust::count_if(thrust::device, fillThrust, fillThrust + sliceNbCellSES, is_notempty());
                                                       ^

CudaSurf.cu(648): error: namespace "thrust" has no member "device"
                  unsigned int notEmptyCells = thrust::count_if(thrust::device, fillThrust, fillThrust + sliceNbCellSES, is_notempty());
                                                                        ^

6 errors detected in the compilation of "CudaSurf.cu".
make: *** [Makefile:21: CudaSurf.o] Error 2

My Cuda info: 
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.98                 Driver Version: 535.98       CUDA Version: 12.2     |
|-----------------------------------------+----------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |         Memory-Usage | GPU-Util  Compute M. |
|                                         |                      |               MIG M. |
|=========================================+======================+======================|
|   0  NVIDIA GeForce RTX 4090        Off | 00000000:01:00.0  On |                  Off |
|  0%   38C    P8              17W / 450W |   1929MiB / 24564MiB |      4%      Default |
|                                         |                      |                  N/A |
+-----------------------------------------+----------------------+----------------------+

from quickses.

miemiemmmm commented on August 18, 2024

Could it be the case that there is no enough shifting of the voxels which causes the truncation of the surface?
There is also some implicit problem with some specific values of grid spacing.

./QuickSES -i 4bso.pdb -o 4bso.obj -v 0.24

Look like a truncation
./QuickSES -i 4bso.pdb -o 4bso.obj -v 0.22

With spacing at 0.2 everything looks fine. 
./QuickSES -i 4bso.pdb -o 4bso.obj -v 0.20

from quickses.

nezix commented on August 18, 2024

I need a bit of time to work on this. Thanks for your report

from quickses.

nezix commented on August 18, 2024

I fixed the compilation errors on recent CUDA versions.
Also, I investigated the issue with some values of the voxel size, the problem comes from the step after the Marching Cubes algorithm, welding vertices.
I am currently trying to fix it

from quickses.

nezix commented on August 18, 2024

I fixed the issue with welding vertices that caused holes and weird artefacts in the mesh for some values of voxel size.
I didn't manage to reproduce the truncation issue on my machine though

from quickses.

Two potential problem I encountered during using the code. about quickses HOT 9 CLOSED

Comments (9)

Related Issues (3)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent