Create annotations ready for rfdetr

- Create `data/reality_rfdetr`
- Split 10% of train into validation data
- Add coco-format annotations
- Add symlink-creation tool
- Update README with new scripts

Author: jaagut

.gitignore

@@ -2,6 +2,7 @@
 /data/**/*
 !/data/**/
 !/data/**/annotations.yaml
+!/data/**/_annotations.coco.json
 
 *.pyc
 __pycache__

README.md

@@ -293,6 +293,14 @@ Creates segmentations for balls and goalposts and field edges.
 
 Splits the dataset into a training and a test set.
 
+#### `split_train_valid.py`
+
+Splits the training dataset into a training and a validation set.
+
+#### `create_symlinks.py`
+
+Creates symlinks for image- and segmentation-files in the data directory to some stored location.
+
 #### `base_footprint_annotator.py`
 
 This script can be used to annotate and verify robot base footprints.
@@ -301,7 +309,6 @@ This script can be used to annotate and verify robot base footprints.
 
 This script can be used to visualize robot base footprint annotations.
 
-
 ### Variational Autoencoder
 
 The variational autoencoder, we have used, is based on [noctrog's conv-vae](https://github.com/noctrog/conv-vae).