dhodcz2
commented
1 year ago Hello, To run the model on your own data, please make sure you have orthorectified images for your region of interest that meet the requirements. Specifically, be sure your images are in XYZ (Slippy) format. The library currently does not accept other formats. https://wiki.openstreetmap.org/wiki/Slippy_map_tilenames https://chrishewett.com/blog/slippy-tile-explorer/
To see options available to tile2net you may run:
python -m tile2net generate --help
In your case you must specify --input e.g. your/path/to/x/y.png (any format for x/y, such as x_y or y/z works) for the orthorectified image set, --zoom which should be at least 19, e.g. 19 for the XYZ zoom level, and --location for the address or bounding box, e.g. Shanghai Tower or 31.232495096130567, 121.50391211277275, 31.234201427826953, 121.50665869484077. Your command would look something like this:
python -m tile2net generate --input <your/path/to/z/x/y.png> --location 'Shanghai Tower' --zoom 19 | python -m tile2net inference
You may also run the interactive example at inference.ipynb if you are familiar with Jupyter Notebooks.
Keep in mind our model is trained on US cities, and may need retraining for different urban environments such as Shanghai. We would love to hear about any feedback or results you have from China. Thank you for your interest.
Russellexplore
Mary-h86
Hello! I am doing a research project on Shanghai, China, about the walking network in Shanghai. I have a question for you. In the file example.sh you gave me, I applied the location and then generated the polygon segmentation of the walking path and the walking network. However, I would like to use my own image set to generate a walk network by referring to the method you gave me. How can I do this? I specialise in architecture and urban design and do not know anything about programming. I would like to use this program to make a pedestrian network in Shanghai so that it can serve the purpose of urban planning and design.