ytcollector 1.0.7__py3-none-any.whl → 1.0.8__py3-none-any.whl

ytcollector/__init__.py

@@ -3,7 +3,7 @@ SBS Dataset Collector - YouTube 영상 수집 및 YOLO-World 검증 파이프라
 """
 from pathlib import Path
 
-__version__ = "1.0.7"
+__version__ = "1.0.8"
 __author__ = "SBS Dataset Team"
 
 # Package root directory

ytcollector/cli.py

@@ -69,9 +69,9 @@ def run_download(args):
         
         if args.fast:
             from .downloader import download_from_txt_parallel
-            results = download_from_txt_parallel(txt_file, task, base_dir, max_count=args.count)
+            results = download_from_txt_parallel(txt_file, task, base_dir, max_count=args.count, skip_verify=args.skip_verify)
         else:
-            results = download_from_txt(txt_file, task, base_dir, max_count=args.count)
+            results = download_from_txt(txt_file, task, base_dir, max_count=args.count, skip_verify=args.skip_verify)
         
         success_count = sum(1 for r in results if r.get('success'))
         total_success += success_count
@@ -189,6 +189,7 @@ Examples:
     download_parser.add_argument('--task', '-t', required=True, nargs='+', choices=VALID_TASKS, help='One or more tasks (e.g. face tattoo)')
     download_parser.add_argument('--count', '-n', type=int, help='Max videos to collect (default: 1000)')
     download_parser.add_argument('--fast', action='store_true', help='Enable fast parallel downloading')
+    download_parser.add_argument('--skip-verify', '-S', action='store_true', help='Skip YOLO verification and save all clips')
     
     # Download single
     single_parser = subparsers.add_parser('download-single', help='Download single video')
@@ -206,6 +207,7 @@ Examples:
     pipeline_parser = subparsers.add_parser('pipeline', help='Full pipeline')
     pipeline_parser.add_argument('--task', '-t', required=True, nargs='+', choices=VALID_TASKS)
     pipeline_parser.add_argument('--verify', action='store_true')
+    pipeline_parser.add_argument('--skip-verify', '-S', action='store_true', help='Skip verification in download stage')
     
     # List tasks
     subparsers.add_parser('list-tasks', help='List available tasks')

ytcollector/downloader.py

@@ -145,7 +145,8 @@ class VideoDownloader:
         self, 
         url: str, 
         timestamp_min: int, 
-        timestamp_sec: int
+        timestamp_sec: int,
+        skip_verify: bool = False
     ) -> Tuple[Optional[Path], Optional[dict]]:
         """
         특정 타임스탬프 기준으로 ±1:30 클립 다운로드
@@ -207,18 +208,24 @@ class VideoDownloader:
         try:
             self.download_segment(url, start_sec, end_sec, temp_path)
             
-            # 5. YOLO 검증
-            logger.info(f"Verifying content for task: {self.task_type}...")
-            # verifier 모듈 사용하여 검증
-            verify_result = verify_clip(temp_path, self.task_type, self.base_dir)
-            
-            if not verify_result.get('is_valid', False):
-                logger.warning(f"Verification failed: No {self.task_type} detected. Deleting...")
-                if temp_path.exists():
-                    temp_path.unlink()
-                return None, None
+            # 5. YOLO 검증 (skip_verify가 False일 때만 수행)
+            if not skip_verify:
+                logger.info(f"Verifying content for task: {self.task_type}...")
+                # verifier 모듈 사용하여 검증
+                verify_result = verify_clip(temp_path, self.task_type, self.base_dir)
+                
+                logger.info(f"Verification Info - Rate: {verify_result.get('summary', {}).get('detection_rate'):.2%}, Is Valid: {verify_result.get('is_valid')}")
+                
+                if not verify_result.get('is_valid', False):
+                    logger.warning(f"Verification failed: No {self.task_type} detected (Rate: {verify_result.get('summary', {}).get('detection_rate'):.2%}). Deleting...")
+                    if temp_path.exists():
+                        temp_path.unlink()
+                    return None, None
+            else:
+                logger.info(f"Skipping verification for task: {self.task_type} (--skip-verify enabled)")
+                verify_result = {'is_valid': True, 'skipped': True}
             
-            # 6. 검증 통과 -> 최종 저장 (순차적 파일명 생성)
+            # 6. 검증 통과(혹은 건너뜀) -> 최종 저장 (순차적 파일명 생성)
             final_path = get_clip_path(self.base_dir, self.task_type, filename=None)
             
             # 이동 (네트워크 드라이브면 shutil.move 사용)
@@ -276,19 +283,17 @@ def parse_txt_line(line: str) -> Optional[Dict]:
         return None
 
 
-def download_from_txt(txt_path: Path, task_type: str, base_dir: Path = None, max_count: int = None) -> list:
+def download_from_txt(txt_path: Path, task_type: str, base_dir: Path = None, max_count: int = None, skip_verify: bool = False) -> list:
     """TXT 파일에서 다운로드 실행 (순차)"""
-    # 기존 로직을 process_single_item 함수로 분리하여 재사용할 수 있으면 좋겠지만,
-    # 코드 구조상 일단 순차 실행 유지하고 parallel 함수 별도 구현
-    return _process_download_loop(txt_path, task_type, base_dir, parallel=False, max_count=max_count)
+    return _process_download_loop(txt_path, task_type, base_dir, parallel=False, max_count=max_count, skip_verify=skip_verify)
 
 
-def download_from_txt_parallel(txt_path: Path, task_type: str, base_dir: Path = None, max_count: int = None) -> list:
+def download_from_txt_parallel(txt_path: Path, task_type: str, base_dir: Path = None, max_count: int = None, skip_verify: bool = False) -> list:
     """TXT 파일에서 병렬 다운로드 실행 (Fast Mode)"""
-    return _process_download_loop(txt_path, task_type, base_dir, parallel=True, max_count=max_count)
+    return _process_download_loop(txt_path, task_type, base_dir, parallel=True, max_count=max_count, skip_verify=skip_verify)
 
 
-def _process_download_loop(txt_path: Path, task_type: str, base_dir: Path = None, parallel: bool = False, max_count: int = None) -> list:
+def _process_download_loop(txt_path: Path, task_type: str, base_dir: Path = None, parallel: bool = False, max_count: int = None, skip_verify: bool = False) -> list:
     from .config import MAX_VIDEOS_PER_TASK, MAX_WORKERS, REQUEST_DELAY_MIN, REQUEST_DELAY_MAX
     import time
     import random
@@ -346,7 +351,8 @@ def _process_download_loop(txt_path: Path, task_type: str, base_dir: Path = None
             output_path, metadata = downloader.download_clip_at_timestamp(
                 url=data['url'],
                 timestamp_min=data['timestamp_min'],
-                timestamp_sec=data['timestamp_sec']
+                timestamp_sec=data['timestamp_sec'],
+                skip_verify=skip_verify
             )
             
             if output_path is None:
@@ -414,38 +420,51 @@ def _process_download_loop(txt_path: Path, task_type: str, base_dir: Path = None
                 if res.get('status') == 'limit_reached':
                     break
 
-    # --- 2단계: 목표 수량을 못 채웠을 경우 YouTube 검색Fallback ---
-    current_count = downloader.get_saved_video_count()
-    if current_count < limit:
+    # --- 2단계: 목표 수량을 못 채웠을 경우 YouTube 검색Fallback (반복 시도) ---
+    max_search_attempts = 5  # 최대 검색 시도 횟수
+    search_attempt = 0
+    processed_urls = set(data['url'] for data in items) # 이미 처리한 URL 중복 검색 방지
+    
+    while downloader.get_saved_video_count() < limit and search_attempt < max_search_attempts:
+        current_count = downloader.get_saved_video_count()
         remaining = limit - current_count
-        logger.info(f"\nTarget not reached ({current_count}/{limit}). Starting YouTube Search fallback for '{task_type}'...")
         
-        # 검색어: 태스크 이름 (필요시 config에서 태스크별 검색어 별도 지정 가능)
-        search_results = downloader.search_youtube(task_type, max_results=remaining * 2)
+        logger.info(f"\n[Search Attempt {search_attempt+1}] Target not reached ({current_count}/{limit}). Searching YouTube for '{task_type}'...")
+        
+        # 검색어: 태스크 이름
+        # 검색 결과 개수를 점진적으로 늘리거나 조절 가능
+        search_results = downloader.search_youtube(task_type, max_results=min(100, remaining * 5))
         
         if not search_results:
-            logger.warning("No search results found.")
-            return results
+            logger.warning("No more search results found.")
+            break
         
-        # 검색 결과는 타임스탬프 정보가 없으므로, 기본적으로 영상의 1:00 지점 혹은 0:00 지점을 시도
-        # 여기서는 영상의 대략 1분 지점(영상이 짧으면 0)을 타겟으로 시도해봄
+        # 새로운 URL만 필터링
+        new_entries = [e for e in search_results if e['url'] not in processed_urls]
+        if not new_entries:
+            logger.info("No new unique videos found in this search attempt.")
+            search_attempt += 1
+            continue
+            
         search_items = []
-        for entry in search_results:
+        for entry in new_entries:
+            processed_urls.add(entry['url'])
+            # 검색 결과는 타임스탬프 정보가 없으므로 여러 지점 시도 가능 (현재는 1분 지점 고정)
             search_items.append({
                 'task_type': task_type,
                 'url': entry['url'],
-                'timestamp_min': 1,  # 1분 지점 샘플링 시도
+                'timestamp_min': 1, 
                 'timestamp_sec': 0,
                 'description': f"Auto-searched: {entry['title']}"
             })
             
-        logger.info(f"Processing {len(search_items)} search results...")
+        logger.info(f"Processing {len(search_items)} new search results...")
         
         if parallel:
             with ThreadPoolExecutor(max_workers=MAX_WORKERS) as executor:
                 futures = [executor.submit(process_item, item) for item in search_items]
                 from tqdm import tqdm
-                for future in tqdm(as_completed(futures), total=len(search_items), desc="Search Fallback"):
+                for future in tqdm(as_completed(futures), total=len(search_items), desc=f"Search Fallback #{search_attempt+1}"):
                     try:
                         res = future.result()
                         results.append(res)
@@ -462,5 +481,13 @@ def _process_download_loop(txt_path: Path, task_type: str, base_dir: Path = None
                 results.append(res)
                 if res.get('status') == 'limit_reached':
                     break
+        
+        search_attempt += 1
+        
+    final_count = downloader.get_saved_video_count()
+    if final_count < limit:
+        logger.warning(f"Finished search attempts. Final count: {final_count}/{limit}")
+    else:
+        logger.info(f"Successfully reached target count: {final_count}/{limit}")
     
     return results

ytcollector/verifier.py

@@ -136,7 +136,7 @@ class YOLOWorldVerifier:
             'frame_results': frame_results,
             'verified_at': datetime.now().isoformat(),
             'model': self.model_name,
-            'is_valid': detection_rate > 0.1,
+            'is_valid': detection_rate > 0.01,  # 1% 이상 탐지되면 유효한 것으로 간주 (기존 10%에서 하향)
         }
         
         logger.info(

{ytcollector-1.0.7.dist-info → ytcollector-1.0.8.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ytcollector
-Version: 1.0.7
+Version: 1.0.8
 Summary: SBS 데이터셋 수집기
 Requires-Python: >=3.9
 Description-Content-Type: text/markdown

ytcollector-1.0.8.dist-info/RECORD

@@ -0,0 +1,12 @@
+config/settings.py,sha256=RcK41kaUC0zam5SsdXfb7u_qjM_TlJDa0a8hC_MGacQ,1197
+ytcollector/__init__.py,sha256=Alq_JsvEi2RZusFB7-AAhzNnMFyBbHNU6NJR_AxDgs4,365
+ytcollector/cli.py,sha256=Fg8GA7gQu3s26XgUg1NBEZuNKQpZLRrnp2s5bqhed8M,8636
+ytcollector/config.py,sha256=ez9flxTbjmdiJB7_IYivWd9xaRfJ8CLBhPYRedLi8Mk,2323
+ytcollector/downloader.py,sha256=QrqwG7PpoJZymKnOOEj1aqZudE6qY3AV4fiuo1anoGk,19901
+ytcollector/utils.py,sha256=gInDx6adV-SfQ2SH5_i8w1gvYL-Nsmz1e1W__gCdVH8,4654
+ytcollector/verifier.py,sha256=OHkAyUF3J6wFqKSa7RYT9Z6-W_lwTXarBEF9xqxNthA,6366
+ytcollector-1.0.8.dist-info/METADATA,sha256=a0t8IXj-xlqbsy_D5EBNemVRi-Z52gJAmu2VJlep0Cc,3727
+ytcollector-1.0.8.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+ytcollector-1.0.8.dist-info/entry_points.txt,sha256=PoanZbxogGnV4tLcZZkla0Yh7OvPtqcukDYr563w5RA,53
+ytcollector-1.0.8.dist-info/top_level.txt,sha256=TVfBZHJgYRfSSTgLJELvOoMA55qR8kWuxtiIaItwzIQ,19
+ytcollector-1.0.8.dist-info/RECORD,,

ytcollector-1.0.7.dist-info/RECORD

@@ -1,12 +0,0 @@
-config/settings.py,sha256=RcK41kaUC0zam5SsdXfb7u_qjM_TlJDa0a8hC_MGacQ,1197
-ytcollector/__init__.py,sha256=tMCSB_dqIAY-1jVYkIrI1PvRYTnL1EkofRYGKh1uN24,365
-ytcollector/cli.py,sha256=meCnT3cMBF15AkHpPeUdCFjL6WHZI-UB_F1echeU6is,8328
-ytcollector/config.py,sha256=ez9flxTbjmdiJB7_IYivWd9xaRfJ8CLBhPYRedLi8Mk,2323
-ytcollector/downloader.py,sha256=ZYCz2oQyVmr9PQG3gVIH17KlZgqomCnNd06n1TyMSR8,18481
-ytcollector/utils.py,sha256=gInDx6adV-SfQ2SH5_i8w1gvYL-Nsmz1e1W__gCdVH8,4654
-ytcollector/verifier.py,sha256=8Nn3b6fTQYxCGPt01kJMDSZ2hy8gk54deSayOpBuY48,6286
-ytcollector-1.0.7.dist-info/METADATA,sha256=79KVtg4U_lrLn-UHTJTugSdggrRdg3kKN5Usqc6kcTM,3727
-ytcollector-1.0.7.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
-ytcollector-1.0.7.dist-info/entry_points.txt,sha256=PoanZbxogGnV4tLcZZkla0Yh7OvPtqcukDYr563w5RA,53
-ytcollector-1.0.7.dist-info/top_level.txt,sha256=TVfBZHJgYRfSSTgLJELvOoMA55qR8kWuxtiIaItwzIQ,19
-ytcollector-1.0.7.dist-info/RECORD,,