Add support to process full areas. Enable batch downloading and inference. (#4)

* Add split_area_into_lat_lon_centers.

* enh(download_tiles): Speeduo download process.

Use ThreadPoolExecutor.
Drop unnecessary usage of PIL for saving.

* feat(demo): Add `run_inference_area`. Rename `run_inference` to `run_inference_point`.

* test ThreadPoolExecutor

* try batch prediction

* Fix last batch call. Expose batch_size. Defaullt to 32.

* Set defaults in notebooks and app

* Update README to show process_area

Author: daavoo

README.md

@@ -38,7 +38,7 @@ Get started right away finding swimming pools and contributing them to OpenStree
 
 | Google Colab | HuggingFace Spaces  | GitHub Codespaces |
 | -------------| ------------------- | ----------------- |
-| [![Run Inference](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/mozilla-ai/osm-ai-helper/blob/main/demo/run_inference.ipynb) | [![Try on Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Try%20on-Spaces-blue)](https://huggingface.co/spaces/mozilla-ai/osm-ai-helper) | [![Try on Codespaces](https://github.com/codespaces/badge.svg)](https://github.com/codespaces/new?hide_repo_select=true&ref=main&repo=928839264&skip_quickstart=true&machine=standardLinux32gb) |
+| Process full Area <br>[![Inference Area](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/mozilla-ai/osm-ai-helper/blob/main/demo/run_inference_area.ipynb) <br> Around a Point <br> [![Inference point](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/mozilla-ai/osm-ai-helper/blob/main/demo/run_inference_point.ipynb) | [![Try on Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Try%20on-Spaces-blue)](https://huggingface.co/spaces/mozilla-ai/osm-ai-helper) | [![Try on Codespaces](https://github.com/codespaces/badge.svg)](https://github.com/codespaces/new?hide_repo_select=true&ref=main&repo=928839264&skip_quickstart=true&machine=standardLinux32gb) |
 
 You can also create your own dataset and finetune a new model for a different use case:
 

demo/app.py

@@ -43,7 +43,7 @@ def __init__(self):
 
 
 @st.fragment
-def inference(lat_lon, margin):
+def inference(lat_lon):
     with st.spinner("Downloading model..."):
         hf_hub_download(
             "mozilla-ai/swimming-pool-detector",
@@ -56,8 +56,9 @@ def inference(lat_lon, margin):
             yolo_model_file="models/model.pt",
             output_dir="/tmp/results",
             lat_lon=lat_lon,
-            margin=margin,
+            margin=2,
             save_full_images=False,
+            batch_size=64,
         )
     return output_path, existing, new
 
@@ -97,7 +98,7 @@ def upload_results(output_path):
         "The results will be uploaded using the [osm-ai-helper](https://www.openstreetmap.org/user/osm-ai-helper) profile."
     )
     st.markdown(
-        "You can check the [Colab Notebook](ttps://colab.research.google.com/github/mozilla-ai/osm-ai-helper/blob/main/demo/run_inference.ipynb)"
+        "You can check the [Colab Notebook](ttps://colab.research.google.com/github/mozilla-ai/osm-ai-helper/blob/main/demo/run_inference_point.ipynb)"
         " and the [Authorization Guide](https://mozilla-ai.github.io/osm-ai-helper/authorization)"
         " to contribute with your own OpenStreetMap account."
     )
@@ -152,7 +153,7 @@ def upload_results(output_path):
 if st.button("Run Inference") and lat_lon:
     lat, lon = lat_lon.split(",")
     output_path, existing, new = inference(
-        lat_lon=(float(lat.strip()), float(lon.strip())), margin=2
+        lat_lon=(float(lat.strip()), float(lon.strip()))
     )
 
     st.info(f"Found {len(existing)} swimming pools already in OpenStreetMaps.")