--- a
+++ b/scripts/annotation-generation-2d.py
@@ -0,0 +1,65 @@
+import json
+import os.path as osp
+from glob import glob
+
+import pandas as pd
+
+# 1. N - Normal
+# 2. V - PVC (Premature ventricular contraction)
+# 3. \ - PAB (Paced beat)
+# 4. R - RBB (Right bundle branch)
+# 5. L - LBB (Left bundle branch)
+# 6. A - APB (Atrial premature beat)
+# 7. ! - AFW (Ventricular flutter wave)
+# 8. E - VEB (Ventricular escape beat)
+
+classes = ["N", "V", "\\", "R", "L", "A", "!", "E"]
+lead = "MLII"
+extension = "png"  # or `npy` for 1D
+data_path = osp.abspath("../data/*/*/*/*/*.{}".format(extension))
+val_size = 0.1  # [0, 1]
+
+output_path = "/".join(data_path.split("/")[:-5])
+random_state = 7
+
+if __name__ == "__main__":
+    dataset = []
+    files = glob(data_path)
+
+    for file in glob(data_path):
+        *_, name, lead, label, filename = file.split("/")
+        dataset.append(
+            {
+                "name": name,
+                "lead": lead,
+                "label": label,
+                "filename": osp.splitext(filename)[0],
+                "path": file,
+            },
+        )
+
+    data = pd.DataFrame(dataset)
+    data = data[data["lead"] == lead]
+    data = data[data["label"].isin(classes)]
+    data = data.sample(frac=1, random_state=random_state)
+
+    val_ids = []
+    for cl in classes:
+        val_ids.extend(
+            data[data["label"] == cl]
+            .sample(frac=val_size, random_state=random_state)
+            .index,
+        )
+
+    val = data.loc[val_ids, :]
+    train = data[~data.index.isin(val.index)]
+
+    train.to_json(osp.join(output_path, "train.json"), orient="records")
+    val.to_json(osp.join(output_path, "val.json"), orient="records")
+
+    d = {}
+    for label in train.label.unique():
+        d[label] = len(d)
+
+    with open(osp.join(output_path, "class-mapper.json"), "w") as file:
+        file.write(json.dumps(d, indent=1))