From 69d35ee5bf1d9a818f437c19636ce3d3ac770c9d Mon Sep 17 00:00:00 2001
From: Alessandro Mondin <alessandro.mondin@ittention.com>
Date: Thu, 9 Nov 2023 10:55:20 +0100
Subject: [PATCH 1/2] feat: export DHF1K to UniSal format

---
 unisal/export_dhf1k.py | 67 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 67 insertions(+)
 create mode 100644 unisal/export_dhf1k.py

diff --git a/unisal/export_dhf1k.py b/unisal/export_dhf1k.py
new file mode 100644
index 0000000..0207f4e
--- /dev/null
+++ b/unisal/export_dhf1k.py
@@ -0,0 +1,67 @@
+"""
+Script used to export DHF1K from original form (after downloading + unar files) to a directory
+compatible with this repository
+"""
+import argparse
+from pathlib import Path
+from shutil import copytree
+
+import cv2
+
+DHF1K_PATH = "/Users/alessandro/datasets/DHF1K"
+
+
+def extract_frames(video_path, frames_dir):
+    vidcap = cv2.VideoCapture(str(video_path))
+    success, image = vidcap.read()
+    count = 0
+    while success:
+        frame_filename = f"{count:04d}.png"
+        cv2.imwrite(str(frames_dir / frame_filename), image)  # Save frame as PNG file
+        success, image = vidcap.read()
+        count += 1
+    vidcap.release()
+
+
+def convert_dhf1k_to_unisal_format(dhf1k_path: str):
+    dhf1k_path = Path(dhf1k_path)
+    dhf1k_unisal_path = Path(__file__).parent.parent / "data" / "DHF1K"
+    dhf1k_unisal_path.mkdir(parents=True, exist_ok=True)
+
+    dhf1k_path_annot = dhf1k_path / "annotation"
+    # only for debugging purposes
+    if (dhf1k_unisal_path / "annotation").exists() is False:
+        copytree(dhf1k_path_annot, dhf1k_unisal_path / "annotation")
+
+    for idx, video in enumerate(sorted((dhf1k_path / "video").glob("*.AVI"))):
+        # original videos in DHF1K are named like 001.AVI, OO2.AVI
+        # but the annotations are in the format like 0001, 0002, 0003.
+        # Hence we need to manually add a "0" to the video naming to
+        # check if the folder already exists within the "annotations"
+        # subfolder. The reason why we need to do this check is
+        # because unisal expects all the images extracted from videos
+        # to be present within the "annotation" subfolder. This folder
+        # that has been copied from the original DHF1K structure however
+        # contains only 700 folders that correspond to the ones that include
+        # label, the remaining 300 folder used for testing are inserted thanks
+        # to this check
+
+        # naming is like 0001, 0999 ... 1000 ... hence for the last folder we
+        # do not add the extra 0
+        annot_vs_video = 0 if idx != 999 else ""
+        video_folder_name = f'{str(annot_vs_video)}{video.parts[-1].replace(".AVI", "")}'
+        video_subfolder = Path(dhf1k_unisal_path, "annotation", video_folder_name, "images")
+        
+        # creates folders form 701 to 1000 and "images" subfolders
+        if video_subfolder.exists() is False:
+            video_subfolder.mkdir(parents=True, exist_ok=True)
+
+        extract_frames(video_path=video, frames_dir=video_subfolder)
+
+
+def main():
+    convert_dhf1k_to_unisal_format(DHF1K_PATH)
+
+
+if __name__ == "__main__":
+    main()

From 7869a4f6cc0f142624099947bc516bbe36ade802 Mon Sep 17 00:00:00 2001
From: Alessandro Mondin <alessandro.mondin@ittention.com>
Date: Mon, 27 Nov 2023 17:01:32 +0100
Subject: [PATCH 2/2] feat: export dhf1k to format required by UniSal

---
 unisal/export_dhf1k.py | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/unisal/export_dhf1k.py b/unisal/export_dhf1k.py
index 0207f4e..96e8e7f 100644
--- a/unisal/export_dhf1k.py
+++ b/unisal/export_dhf1k.py
@@ -2,19 +2,20 @@
 Script used to export DHF1K from original form (after downloading + unar files) to a directory
 compatible with this repository
 """
-import argparse
 from pathlib import Path
 from shutil import copytree
 
 import cv2
 
-DHF1K_PATH = "/Users/alessandro/datasets/DHF1K"
+# path to https://drive.google.com/drive/folders/1sW0tf9RQMO4RR7SyKhU8Kmbm4jwkFGpQ
+# DHF1K_PATH = "/Path/To/Dataset/DHF1K"
+DHF1K_PATH = "/Users/Alessandro/Datasets/DHF1K"
 
 
 def extract_frames(video_path, frames_dir):
     vidcap = cv2.VideoCapture(str(video_path))
     success, image = vidcap.read()
-    count = 0
+    count = 1
     while success:
         frame_filename = f"{count:04d}.png"
         cv2.imwrite(str(frames_dir / frame_filename), image)  # Save frame as PNG file