3 ms·
And rather unfortunately has a lack of documentation. Many of the goals here has been targets of more traditional gis image processing algorithms for quite a wh
by anakaine 5y ago
And rather unfortunately has a lack of documentation. Many of the goals here has been targets of more traditional gis image processing algorithms for quite a while. There has been very active work in that space for years with a good number of year on year improvements. Additionally, footprint detection and regularisation isn't particularly easy if you want to go a good job, and roof material can be quite difficult wven from 13 band imagery.
Putting this in BlenderGIS seems like reinventing a good many wheels to be honest.
- ivaneres 5y agoDeveloper of the repo here. To be honest, I agree with most of these points - this project shares a goal similar to large companies like https://blackshark.ai/ https://blackshark.ai/ , so I don't expect to be able to achieve results of a similar standard. If anything, this project was just meant to be an experiment to play around with computer vision and Blender in my free time, I never expected it to receive attention, and especially not this early on. That being said, seeing the interest around this idea, I will make an effort to translate my research to a viable product. As you said, the current readme is quite lackluster, mostly as I haven't yet figured out how everything will work - I'll be updating it as the details become more clear. w.r.t the BlenderGIS point, this is something I am yet to decide on, so I'll see what the most suitable platform will be when I get to that stage of the project.
- danuker 5y agoNo worries. I was criticizing OP; it is not ready to be reviewed or tried out for now. I also had a goal in high school to make a GTA-style 3D game of my home town. Needless to say, I did not succeed. I gave up on it, but I learned a lot about 3D modeling, and motivated me to learn Python. I wish you good luck!
- ivaneres 5y agoThank you!