Built on your original semantic segmentation work with these additions:
| """Example: TiledInferenceCallback for large-scale geospatial inference. | |
| This script demonstrates how to use TorchGeo's TiledInferenceCallback to run | |
| inference on large rasters that don't fit in memory. It handles: | |
| - Tiled processing with configurable patch size and overlap | |
| - Weighted blending at patch boundaries (cosine/linear) | |
| - Automatic geospatial metadata preservation | |
| - Chunked GeoTIFF output for large results | |
| Usage: |
| # This webapp is based on Healthy or Not (https://github.com/nikhilno1/healthy-or-not) -> Thanks! | |
| from starlette.applications import Starlette | |
| from starlette.responses import JSONResponse, HTMLResponse, RedirectResponse | |
| from fastai import * | |
| from fastai.vision import * | |
| import torch | |
| from io import BytesIO | |
| import uvicorn | |
| import aiohttp |