init commit

dataset.py

@@ -0,0 +1,75 @@
+from datasets import load_dataset
+import tqdm
+from xml.etree.ElementTree import ParseError
+from youtube_transcript_api._errors import TranscriptsDisabled
+import json
+from youtube_transcript_api import YouTubeTranscriptApi
+
+dataset = load_dataset("OpenGVLab/InternVid-Full")
+
+video_id = set()
+for idx, row in tqdm.tqdm(enumerate(dataset['train'])):
+    video_id.add(row['YoutubeID'])
+    
+    if len(video_id) == 50000:
+        break
+
+video_id_list = list(video_id)
+
+def create_transcript(video_id, transcripts):
+    candidate = {}
+    candidate["video_id"] = video_id
+    candidate["transcripts"] = transcripts.to_raw_data()
+    return candidate
+
+
+
+res = []
+prev = ""
+
+for video_id in tqdm.tqdm(video_id_list):
+    if video_id == prev:
+        print("sth is wrong")
+    retry = False
+    while True:
+        try:
+            ytt_api = YouTubeTranscriptApi()
+            transcript_list = ytt_api.list(video_id)
+            ok = False
+            for transcript in transcript_list:
+                if transcript.language == "en":
+                    data = transcript.fetch()
+                    print(f"English is available for this video: {video_id}")
+                    res.append(create_transcript(video_id, data))
+                    ok = True
+                    break
+            if not ok:
+                for transcript in transcript_list:
+                    if transcript.is_translatable:
+                        data = transcript.translate('en').fetch()
+                        print(f"[TRAN] Translated to English for {video_id}")
+                        res.append(create_transcript(video_id, data))
+                        ok = True
+                        break
+            if not ok:
+                print(f"There is no English version for the video")
+            break
+        except ParseError as e:
+            if not retry:
+                print(f"[Retry] XML ParseError for {video_id}: {e}. Retrying...")
+                retry = True
+            continue
+        except TranscriptsDisabled as e:
+            print(f"[Skip ] TranscriptsDisabled for {video_id}. Skipping.")
+            break
+        except Exception as e:
+            # print( print(f"[Error] Unexpected error for {video_id}: {e}. Skipping."))
+            break
+    prev = video_id
+    # if candidate:
+    #     res.append(candidate)
+
+print(len(res))
+
+with open("InternVid_1.json", "w") as f:
+    json.dump(res, f)