Comments (11)
The issue should be fixed by the latest commit. You can pull the latest version and try again. Sorry for the caused trouble.
To provide some details, the X-Y plane in the ARKit coordinates should be moved downwards a little to meet the training settings on ScanNet.Hello! Latest commit fixed an error, but the results are still unrecognizable.
Should results on demo data look like this? https://drive.google.com/file/d/1g-woNtSl8duVeV2IETtxlDB-PjVmYo4a/view?usp=sharingIt appears that the overall shape is correct and the problem is on the shading mode in MeshLab. You can try to use this shading configuration in MeshLab:
The released demo data is the data used in the header video on the project page. The corresponding mesh is also hosted on Skechfab.
Turns out I loaded all the generated meshes at once. Now everything's fine, thank you for the quick response!
from neuralrecon.
I reloaded demo data, and it worked fine after that.
from neuralrecon.
@JiamingSuen Thank you for looking into this super fast. It works great, and I was able to test it on my own captured data as well. Awesome work!
from neuralrecon.
I get the same error, for all the incremental saves and final save, with or without visualization
from neuralrecon.
It is possible to get images and intermediate meshes by adding a condition before save_mesh_scene.
if "scene_tsdf" in outputs:
save_mesh_scene(outputs, sample, epoch_idx)
However, the result is completely unrecognizable.
from neuralrecon.
The issue should be fixed by the latest commit. You can pull the latest version and try again. Sorry for the caused trouble.
To provide some details, the X-Y plane in the ARKit coordinates should be moved downwards a little to meet the training settings on ScanNet.
UPDATE: Please notice that the data preprocessing needs to be run again after pulling the latest code, which can be achieved by simply deleting SyncedPoses.txt
in the data folder.
from neuralrecon.
The issue should be fixed by the latest commit. You can pull the latest version and try again. Sorry for the caused trouble.
To provide some details, the X-Y plane in the ARKit coordinates should be moved downwards a little to meet the training settings on ScanNet.
Hello! Latest commit fixed an error, but the results are still unrecognizable.
Should results on demo data look like this? https://drive.google.com/file/d/1g-woNtSl8duVeV2IETtxlDB-PjVmYo4a/view?usp=sharing
from neuralrecon.
The issue should be fixed by the latest commit. You can pull the latest version and try again. Sorry for the caused trouble.
To provide some details, the X-Y plane in the ARKit coordinates should be moved downwards a little to meet the training settings on ScanNet.Hello! Latest commit fixed an error, but the results are still unrecognizable.
Should results on demo data look like this? https://drive.google.com/file/d/1g-woNtSl8duVeV2IETtxlDB-PjVmYo4a/view?usp=sharing
It appears that the overall shape is correct and the problem is on the shading mode in MeshLab. You can try to use this shading configuration in MeshLab:
The released demo data is the data used in the header video on the project page. The corresponding mesh is also hosted on Skechfab.
from neuralrecon.
@ftasse Please confirm that you have also been able to run the demo with the latest commit, thanks!
from neuralrecon.
@JiamingSuen Thank you for looking into this super fast. It works great, and I was able to test it on my own captured data as well. Awesome work!
Thanks for your reply (and also for raising this issue)! Glad that you like our work!
from neuralrecon.
The issue should be fixed by the latest commit. You can pull the latest version and try again. Sorry for the caused trouble.
To provide some details, the X-Y plane in the ARKit coordinates should be moved downwards a little to meet the training settings on ScanNet.
UPDATE: Please notice that the data preprocessing needs to be run again after pulling the latest code, which can be achieved by simply deleting
SyncedPoses.txt
in the data folder.
Hello, thank you for your great work.
Could you explain that why the ARKit coordinates should be moved downwards a little.
BTW, when we run NeuralRecon on our own data, it occurred no valid points: scale 1/2
like this issue.
Whether should we pre-process our own data to meet some kind of
training settings on ScanNet ?
Could you provide the training setting in detail?
from neuralrecon.
Related Issues (20)
- Error when trying to convert to onnx HOT 1
- Some question about the codes
- Google Colab Implementation
- Google Colab Implementation
- generate_gt.py memory issue HOT 2
- generate_gt.py stuck in different locations
- Building the dataset according to given structure
- scannet.py FileNotFoundError HOT 1
- generate_gt.py got cuda not detected error HOT 1
- skimage marching_cubes_lewiner error HOT 2
- What should I do to perform 3D reconstruction from the images I have acquired?
- ImportError: cannot import name 'PointTensor' from 'torchsparse.tensor' HOT 2
- Export the ONNX newspaper too many indices for tensor of dimension 0
- Why inference on ScanNet test-set didn't test the scene 806?
- The SyncBatchNorm function will stop, when one of multi-GPU run into "no valid points". How to solve it?
- [Question] about: 3D reconstruction from image slides
- 7-Scenes dataset processing
- Problem in running generate_gt.py
- Project server configuration requirements
- as for cuda=11.8,update env.yml HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from neuralrecon.