ywyue / roomformer Goto Github PK
View Code? Open in Web Editor NEW[CVPR 2023] RoomFormer: Two-level Queries for Single-stage Floorplan Reconstruction
Home Page: https://ywyue.github.io/RoomFormer/
License: MIT License
[CVPR 2023] RoomFormer: Two-level Queries for Single-stage Floorplan Reconstruction
Home Page: https://ywyue.github.io/RoomFormer/
License: MIT License
When I trained directly with the processed Structured3D dataset, the final training result seemed to be wrong. I don't know what I did wrong that led to this.
When training, I used the command:
python main.py --dataset_name=stru3d --dataset_root=dataru3d --num_queries=800 --num_polys=20 --semantic_classes=-1 --job_name=train_stru3d --num_workers=0
The relevant indicators obtained after the training are as follows:
Averaged stats: room_prec: 0.0000 (0.1604) room_rec: 0.0000 (0.1415) corner_prec: 0.0000 (0.0777) corner_rec: 0.0000 (0.0607) angles_prec: 0.0000 (0.0626) angles_rec: 0.0000 (0.0500) loss: 0.8760 (0.8640) loss_ce: 0.1437 (0.1420) loss_coords: 0.3569 (0.3627) loss_raster: 0.3638 (0.3593) loss_ce_unscaled: 0.0719 (0.0710) loss_coords_unscaled: 0.0714 (0.0725) loss_raster_unscaled: 0.3638 (0.3593) cardinality_error_unscaled: 9.7000 (9.8400)
Thank you for your open source work. When I tried to use my own three-channel image data set, it seemed that I could not train, there were no bugs in the training, and the loss was constantly decreasing, but the accuracy of the valuation set was always 0. Do you know the possible reasons for this?
Hi, thanks for your beautiful work. How can I visualize the results from the processed scenecad data?
Where's the code to test this out?
Hi Team,
Thank you for the code and pre-trained model.
But when I am trying to load model to GPU memory I am getting "Segmentation fault (core dumped)" message. I am currently using 16 GB GPU machine with the same environment as the repository suggested. Also tried with 24 GB machine but still getting the same error. What could be the reason for this? And if anyone has any solution kindly suggest the same.
This paper is great ! Can you publish the code? I will thank you very much!
Hi yuewen, thanks for your perfect work and congratulations that roomformer is accepted by CVPR!
However, I have some problems trainning roomformer to get the good resluts, expecially with the angle precision and angle recall, and about 2 points lower than the results in the paper in the six evaluation metrics. Besides, in the code, the checkpoint is saved per 20 epoch, i wonder if there are any better way to save the checkpoints?
Can you give me some advice on how to train the model?
Thanks a lot if you can help me!
How to get the semantic-rich floorplan result?
Hi~ ywyue, thank you for your work.
When I execute the code "generate_point_cloud_stru3d.py", I get the following error.
What should be added is that some files in the series of compressed packages named "Structured3D_panorama" that I downloaded were corrupted, so I had to delete part of the folder whose name format is "scene_xxxxx". I don't know if this will have any impact. At the time of the error, the code has processed all four of the zipped data.
77%|████████████████████████████████████████████████████████████████████████████████████████████████████Pointcloud size: 1036436█████████████████████████▋ | 152/198 [3:36:07<47:42, 62.23s/it]
77%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████▏ | 153/198 [3:38:25<1:04:14, 85.65s/it]
26%|███████████████████▍ | 5/19 [28:54:28<80:56:31, 20813.65s/it]
Traceback (most recent call last):
File "generate_point_cloud_stru3d.py", line 29, in
main(config())
File "generate_point_cloud_stru3d.py", line 22, in main
reader = PointCloudReaderPanorama(scene_path, random_level=0, generate_color=True, generate_normal=False)
File "E:\YangMeiQi\git\RoomFormer\data_preprocess\stru3d\PointCloudReaderPanorama.py", line 25, in init
self.point_cloud = self.generate_point_cloud(self.random_level, color=self.generate_color, normal=self.generate_normal)
File "E:\YangMeiQi\git\RoomFormer\data_preprocess\stru3d\PointCloudReaderPanorama.py", line 75, in generate_point_cloud
coords[:,:2] = np.round(coords[:,:2] / 10) * 10.
IndexError: too many indices for array: array is 1-dimensional, but 2 were indexed
Hi, ywyue!
Thank you for your wonderful work.
I tried to train on the Structured3D dataset, however, the training got stuck midway without any error being reported.
I tried to set --num_workers=0, but the problem hasn't been resolved.
I've tried terminating and resuming training multiple times, but the epoch at which it gets stuck varies each time. Do you have any suggestions for a solution?
I'm using 1.9.0+cu111 and running the main.py in a WSL2 Ubuntu 20.04
ModuleNotFoundError:No module named 'detectron2.data'
Could you please provide the corresponding code?
An error occurred while trying to convert the stru3d dataset to a point cloud. It seems that the dataset I downloaded did not have the same directory as in the code after extracting it.
error:
Creating point cloud from perspective views...
0%| | 0/3500 [00:00<?, ?it/s]
Traceback (most recent call last):
File "data_preprocess/stru3d/generate_point_cloud_stru3d.py", line 29, in
main(config())
File "data_preprocess/stru3d/generate_point_cloud_stru3d.py", line 19, in main
scenes = os.listdir(os.path.join(data_root, part, 'Structured3D'))
FileNotFoundError: [Errno 2] No such file or directory: '/mnt/data2/数据集/Structured3D/Structured3D/scene_02893/Structured3D'
Thanks for the great work!
This is not an issue but more of a question. Can your work and implementation be utilized to detect the rooms and walls directly from a 2D floor plan image? Meaning that, if I bypass the 3D part, can actual 2D floorplan images or pdfs be somehow used as a "Density Map" in your setting?
Some data in Structured3D is lost, resulting in the failure of generating point cloud data. Have you used this tool to convert point clouds?
generate_point_cloud_stru3d.py
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.