Update README.md

README.md

@@ -418,6 +418,10 @@ dataset_info:
   download_size: 4113604095
   dataset_size: 4829359475.84
 ---
+### Load images directly
+You can download images directly from the `images/` directory. The images' names are the ones referenced in the corresponding metadata csv files. 
+
+⚠️ Please note that each line contains a single annotation (single box/polygon), but there can be several runways per image. It means that several lines in the csv file can refer to the same image file.
 
 ### Load with 🤗 Datasets
 You can download this dataset through HuggingFace `datasets` (which requires you to install the extra Vision packages with `pip install datasets[vision]`). Then you will be able to load specific parts of this dataset with:
@@ -425,4 +429,7 @@ You can download this dataset through HuggingFace `datasets` (which requires you
 from datasets import load_dataset
 ds = load_dataset("ORG/REPO_ID", name="flsim")  # or arcgis/bingmaps/ges/xplane
 print(ds)
-```
+```
+⚠️ Please note the following 2 limitations for this method:
+- In the resulting parquet file, the image file associated with each annotation was copied directly, meaning this image is duplicated for every line referencing it.
+- The file name is not directly visible in the parquet file, but you can retrieve it by using `cast_column("image", Image(decode=False))`.