Added information about splits and provided metadata

README.md

@@ -16,7 +16,7 @@ dataset_summary: '
 
 
 
-  This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 8000 samples.
+  This is a [FiftyOne](https://github.com/voxel51/fiftyone) dataset with 8,000 samples. The train set consists of 5,288 samples and the test set of 2,712, as marked by FiftyOne sample tags.
 
 
   ## Installation
@@ -87,14 +87,12 @@ session = fo.launch_app(dataset)
 
 ## Dataset Details
 
-
 Fisheye lens provides omnidirectional, wide coverage for using fewer cameras to monitor road intersections, however, with view distortions.
 
 The FishEye8K benchmark dataset for road object detection tasks comprises 157K bounding boxes across five classes (Pedestrian, Bike, Car, Bus, and Truck).
 
 The dataset comprises 8,000 images recorded in 22 videos using 18 fisheye cameras for traffic monitoring in Hsinchu, Taiwan, at resolutions of 1080×1080 and 1280×1280.
 
-
 - **Curated by:** Munkhjargal Gochoo, et. al.
 - **Shared by:** [Harpreet Sahota](https://huggingface.co/harpreetsahota)
 - **Language(s) (NLP):** en
@@ -117,6 +115,8 @@ Object detection
   - T is the time of day indicator (A for afternoon, E for evening, N for night, M for morning).
   - YYY is an arbitrary sequence of digits.
 
+Each FiftyOne sample has two dedicated metadata fields storing the information provided in the filename, **location** for X and **time_of_day** for T. The extraction script was provided by [Daniel Bogdoll](https://github.com/daniel-bogdoll).
+
 ## Dataset Creation
 
 ### Source Data