Closed NourSoltani closed 1 year ago
We are actually in the process of adding a demo script that shows how to download some public data sets, setup the configuration files and run the pipeline. This is in PR #43.
It sounds great! I am looking forward to hearing great news soon and make it run
@mleotta Thank you for your response Mr Leotta. I have another question if you do not mind. In case that I would like to do the inference of the model on another images.. For instance, I would to do a reconstruction for the city where I live, how can I do that? Is it just do simply install data from sentinel etc.. or is there any pre-processing do to before? And If there is, could you guide me how to do so? And thank you for your comprehension. Have a nice day!
Unfortunately data access is the big limiting factor here. This software was designed for high resolution satellite imagery, specifically Maxar WorldView 2 & 3. That imagery is not freely available. So you would need to obtain multiple images of your location with different view angles from Maxar. Sentinel-2 imagery is far lower resolution and doesn't typically have the view angle diversity needed.
@mleotta Thank you for your clarifying Mr Leotta. I really appreciate it! And I would like to apologize for responding late. So, let's say we can provide and buy this resolution of images, is it possible, and easy :), to do the reconstruction for any place we want? And sorry for the multiple questions..
It's certainly possible. How easy it is is a bit subjective. If you have access to multiple WorldView 2 or 3 images from different viewpoints of a site then it should be relatively easy to run. Keep in mind the quality of the results will depend on the quality and amount of input data. The CORE3D sample data sets have 20-30 images, cloud-free, collected from a few view points over only a few years (so minimal scene change). You can probably get something to work with less data, but quality may not be as good. Also we are using the raw unprocessed PAN and MSI data from Maxar. Often when you order imagery it will have already been orthorectified. Orthorectified data is not helpful for 3D modeling. I suggesting testing on the CORE3D data first to see if you can run it.
Thank you Mr @mleotta ! I will download the data you provided as soon as possible, And then if we managed to make the algorithm run, we will see if we can afford the specific data for the parcel we want to reconstruct.. Thank you again!
Hello! I hope you all are doing very well.. I am asking if anyone managed to make the code runs, because I didn't know to set the paths in the input.ini file; for example: the " Imagery source directory", Is there any sample to try on?