Hello,
I would like to know if a dataset exists that would contain lots of different 3D models of buildings along with annotations (church, house, castle, barn, bridge, etc). This dataset would be used to research automatic synthesis of buildings and cities from these exemplars using Deep Neural Networks (DNN) or other approaches.
A good example of this kind of dataset but not dedicated to buildings would be Princeton ShapeNet dataset:
https://shapenet.org/
If something similar would exist for buildings, it would be a great help for my research.
Regards,
Bruno
CAD /CAM solutions will have this (they are all proprietary )
https://www.dlubal.com/en-US/solutions/industries/other/analysis-and-design-software-for-buildings
http://www.iitk.ac.in/nicee/wcee/article/13_1903.pdf
https://www.sema-soft.de/en/software/timber-construction/structural-analysis/
Well , all these are Engineering design ...
Thanks Aparna!
What you suggested is not really what I need in order to study 3D model synthesis. For this purpose, here is what would be needed:
- Only the exterior 3D model of the building, anything inside would be useless (i.e. no real complete CAD construction plans).
- At least a few building categories for which there would be lot (hundreds) of 3D models (house, duplex, multi-apartment buildings, commercial buildings, church, hospital, etc.)
- Ideally, that these model be accompanied by metadata describing it (ex: residential, house, bridge, etc)
A very good example of the type of dataset I'm looking for (w.r.t. size, number of categories, metadata, etc) is really the Princeton ShapeNet dataset (but specific to buildings) :
https://shapenet.org/
Thanks anyway,
Bruno
Sorry !!!!!!!!!!!!
I was wanting to say something and No is what I started with (Sorry )
I mean : Architectural analysis is only to be found in academics .
E.g: paper does this
https://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XL-5-W4/351/2015/isprsarchives-XL-5-W4-351-2015.pdf
Academics prepare them call them dataset ...
If you want to capitalize then getting city data survey or monumental information and build a clean dataset by DCNN.
If something similar would exist for buildings, it would be a great help for my research.
Try to develop a similar one by yourself.
You are a researcher.
Why do you seek others help?
If I help you, Will you include my Name in your publishing works?
Hello Nishad,
As I did not get any positive answer, I will probably develop a very restricted version of what I had in mind, like a few hundred models of 2-3 classes of buildings extracted from Montreal city Open Data.
My project is in the context of a commercial contract and the result of this research can't be shared or published. My time is also limited and I know too well that making something like ShapeNet data set takes huge amount of time which I don't have. By the way, I'm very grateful for the labs that takes the time to do such things for the community.
Finally, I seek others opinion and guidance in a problem domain I'm not familiar with because I have too often seen newcomers in a field lacking humility spending lots of energy for mediocre results just because they did not take the time to get to know their new domain main tools and methods. Also, as I said, my time is limited ant I don't want to reinvent the wheel (which we know already exists)!
Cheers
An interesting paper using another approach:
Article Guided Proceduralization: Optimizing Geometry Processing and...
Can you give a brief explanation of your Project. The abstract, objectives.Then only I can help you.
Sorry Nishad, the details of the project are confidential but my needs are well explained in the original question and the first response I gave to Aparna.
Hello Bruno,
Thanks for referring to our paper. We used architectural models gathered from public sources, with a few internal models. If you are just looking for building geometry with no semantic labels/segmentation/information/etc., then Trimble Warehouse is your friend: https://3dwarehouse.sketchup.com Some of them have tags and metadata, however mining and categorizing them would be a task by itself, because they are mostly not uniform, missing, and subjective. However geometry-wise those are the most detailed, realistic, and diverse buildings.
If you are looking for massive models, another source you can benefit is the following repo. For example the NY model has over 4000 buildings: https://www.3dcadbrowser.com/download.aspx?3dmodel=40454 Again, only geometry and textures, no semantic information. Your best bet would be taking a popular city, download the model from this website, and then cross it with other sources that you can get labels from, like OSM, or open government datasets. There is also this platform for more city science, however I am not sure if they are still active: http://www.urbansim.com/
There are also some small point cloud datasets used in several publications, you need to dig into the projects of the following groups:
https://cvg.ethz.ch/research/
http://www.prs.igp.ethz.ch/research/Source_code_and_datasets.html
https://www.ikg.uni-hannover.de/
I hope this helps!
Ilke
- Badges
- Science topic