add ReadMe

README.md

@@ -1,30 +1,44 @@
----
-Dataset Description
-===================
+# Runway Piano Markings Dataset
 
-This dataset contains 8,000 satellite images of runway piano markings, captured from Google Maps satellite view with a consistent heading. 
+## Overview
+This dataset contains 8,000 satellite images of runway piano markings captured from Google Maps (satellite view). All images share a consistent heading and orientation, making the dataset suitable for detection, localization, and regression tasks involving runway markings.
 
-Dataset Structure:
-------------------
-- `images/` : Contains 8,000 images of size 640x640 pixels.
-    - Images are named using the format:  
-        `{airport_ident}_{runway_ident}.png`  
+## Dataset Structure
+images/
+├── {airport_ident}_{runway_ident}.png
+└── ... (8,000 images, 640×640)
 
-- `labels/` : Contains corresponding `.txt` files for each image.
+labels/
+├── {airport_ident}_{runway_ident}.txt
+└── ... (one label file per image)
 
-Label Format:
--------------
-Each `.txt` file follows the structure:
+- images/: PNG images of size 640 × 640 pixels
+- labels/: Text files containing the corresponding bounding box annotations
 
-    x0,y0,x1,y1
+Each image has exactly one corresponding label file with the same base filename.
 
-where:
-- (x0, y0) represents the top-left corner of the piano.
-- (x1, y1) represents the bottom-right corner of the piano.
+## File Naming Convention
+Images and labels follow the naming scheme:
+{airport_ident}_{runway_ident}
 
-Each label file corresponds to an image in the `images/` folder with the same filename (but with a `.txt` extension).  
+Example:
+LFPG_09L.png
+LFPG_09L.txt
 
+## Annotation Format
+Each label file contains a single line with four comma-separated values:
+x0,y0,x1,y1
 
+Where:
+- (x0, y0): Top-left corner of the runway piano marking
+- (x1, y1): Bottom-right corner of the runway piano marking
 
-license: mit
----
+All coordinates are expressed in pixel space, relative to the top-left corner of the image.
+
+## Notes
+- Each image contains one runway piano marking
+- All runways are oriented in the same direction due to consistent map heading
+- Suitable for object detection, bounding-box regression, and localization tasks
+
+## License
+MIT License