change file style

app.py

@@ -5,9 +5,12 @@ import librosa
 import numpy as np
 import subprocess
 import sys
-from huggingface_hub import snapshot_download
-from pathlib import Path
 import os
+import glob
+from pathlib import Path
+from huggingface_hub import snapshot_download
+
+token = os.getenv("HF_TOKEN")
 
 # Runtime installation of madmom
 def install_madmom():
@@ -45,13 +48,12 @@ except ImportError as e:
         return {
             'matches': [],
             'message': 'madmom dependency not available'
-        } 
-
-token = os.getenv("HF_TOKEN")
+        }
 
 def download_data_from_hub():
     """
     Download covers80 and ml_models folders from Dataset repository
+    (1005_e_4 file is handled by Git LFS in the main repo)
     """
     base_dir = Path(".")
     data_repo_id = "nininigold/music-data"
@@ -59,13 +61,24 @@ def download_data_from_hub():
     folders_to_check = ["covers80", "ml_models"]
     downloaded_folders = {}
     
+    # 1005_e_4 파일은 Git LFS로 이미 있는지 확인
+    lfs_file = base_dir / "1005_e_4"
+    if lfs_file.exists():
+        file_size = lfs_file.stat().st_size / (1024*1024)  # MB
+        print(f"✅ LFS file 1005_e_4: {file_size:.1f} MB")
+        downloaded_folders["1005_e_4"] = str(lfs_file)
+    else:
+        print(f"⚠️ LFS file 1005_e_4 not found")
+        downloaded_folders["1005_e_4"] = None
+    
     # 폴더들이 이미 존재하는지 확인
-    all_exist = all((base_dir / folder).exists() and any((base_dir / folder).iterdir()) 
-                   for folder in folders_to_check)
+    all_folders_exist = all((base_dir / folder).exists() and any((base_dir / folder).iterdir()) 
+                           for folder in folders_to_check)
     
-    if not all_exist:
+    if not all_folders_exist:
         print(f"📥 Downloading data folders from dataset: {data_repo_id}")
         print(f"   This includes covers80 and ml_models folders (~17k+ files each)")
+        print(f"   1005_e_4 file is handled by Git LFS in main repo")
         print(f"   This may take several minutes...")
         
         try:
@@ -110,10 +123,41 @@ def download_data_from_hub():
     
     return downloaded_folders
 
+def find_song_file_by_title(song_title):
+    """
+    covers80 폴더에서 곡 제목으로 mp3 파일을 찾음
+    """
+    covers80_path = Path("covers80")
+    
+    if not covers80_path.exists():
+        return None
+    
+    # 정확한 매치 시도
+    exact_patterns = [
+        f"{song_title}.mp3",
+        f"*{song_title}.mp3",
+        f"{song_title}*.mp3"
+    ]
+    
+    for pattern in exact_patterns:
+        matches = list(covers80_path.glob(pattern))
+        if matches:
+            return str(matches[0])
+    
+    # 부분 매치 시도 (song_title의 일부분들로)
+    song_parts = song_title.replace('_', ' ').split()
+    for part in song_parts:
+        if len(part) > 3:  # 너무 짧은 단어는 제외
+            matches = list(covers80_path.glob(f"*{part}*.mp3"))
+            if matches:
+                return str(matches[0])
+    
+    return None
+
 @spaces.GPU
 def process_audio_for_matching(audio_file):
     """
-    Process the uploaded audio file and return matching results
+    Process the uploaded audio file and return matching results with timestamp playback
     """
     if audio_file is None:
         return """
@@ -127,23 +171,6 @@ def process_audio_for_matching(audio_file):
         # inference 함수 호출
         result = inference(audio_file)
         
-        # result 구조:
-        # {
-        #     'matches': [
-        #         {
-        #             'rank': 1,
-        #             'score': 0.95,
-        #             'song_title': 'Song Name',
-        #             'segment_file': 'path/to/segment.wav',
-        #             'test_time': 23.5,
-        #             'library_time': 45.2,
-        #             'confidence': '95.0%',
-        #             'time_match': 'Input: 23.5s ↔ Library: 45.2s'
-        #         }
-        #     ],
-        #     'message': 'success' or error message
-        # }
-        
         if result.get('message') != 'success':
             return f"""
             <div style="text-align: center; padding: 25px; background: #fff3cd; border-radius: 15px; border: 1px solid #ffeaa7; margin: 10px 0;">
@@ -169,17 +196,51 @@ def process_audio_for_matching(audio_file):
             rank = match.get('rank', 0)
             song_title = match.get('song_title', 'Unknown Song')
             confidence = match.get('confidence', '0%')
-            time_match = match.get('time_match', 'Unknown')
             test_time = match.get('test_time', 0)
             library_time = match.get('library_time', 0)
-            segment_file = match.get('segment_file', '')
             
             # 랭킹에 따른 색상 설정
             rank_colors = {1: '#e74c3c', 2: '#f39c12', 3: '#27ae60'}
             rank_color = rank_colors.get(rank, '#7f8c8d')
             
-            # 세그먼트 파일 정보
-            segment_info = f"Found: {segment_file}" if segment_file else "No segment file found"
+            # 곡 파일 찾기
+            song_file_path = find_song_file_by_title(song_title)
+            
+            # 오디오 플레이어 생성
+            audio_player = ""
+            if song_file_path and os.path.exists(song_file_path):
+                # 상대 경로로 변경
+                rel_song_path = os.path.relpath(song_file_path)
+                audio_player = f"""
+                <div style="margin: 15px 0; padding: 15px; background: #f0f8ff; border-radius: 8px;">
+                    <div style="text-align: center; margin-bottom: 10px;">
+                        <strong style="color: #2c3e50;">🎵 Play from matched timestamp</strong>
+                    </div>
+                    <audio controls preload="metadata" style="width: 100%;">
+                        <source src="file/{rel_song_path}#t={library_time:.1f}" type="audio/mpeg">
+                        <source src="file/{rel_song_path}" type="audio/mpeg">
+                        Your browser does not support the audio element.
+                    </audio>
+                    <div style="text-align: center; margin-top: 8px;">
+                        <button onclick="seekToTime(this.parentElement.previousElementSibling.querySelector('audio'), {library_time})" 
+                                style="background: #3498db; color: white; border: none; padding: 5px 15px; border-radius: 15px; cursor: pointer; font-size: 0.9em;">
+                            ⏭️ Jump to {library_time:.1f}s
+                        </button>
+                    </div>
+                    <p style="font-size: 0.8em; color: #666; text-align: center; margin: 5px 0 0 0;">
+                        Match found at {library_time:.1f}s in the original song
+                    </p>
+                </div>
+                """
+                file_info = f"Found: {os.path.basename(song_file_path)}"
+            else:
+                audio_player = f"""
+                <div style="margin: 10px 0; padding: 10px; background: #fff3cd; border-radius: 8px; text-align: center;">
+                    <p style="color: #856404; margin: 0;">🔍 Song file not found for playback</p>
+                    <p style="font-size: 0.8em; color: #856404; margin: 5px 0 0 0;">Match at {library_time:.1f}s in "{song_title}"</p>
+                </div>
+                """
+                file_info = f"Song file not found: {song_title}"
             
             matches_html += f"""
             <div style="background: #ffffff; border-radius: 12px; padding: 20px; margin: 15px 0; 
@@ -209,8 +270,10 @@ def process_audio_for_matching(audio_file):
                     </div>
                 </div>
                 
+                {audio_player}
+                
                 <div style="font-size: 0.9em; color: #7f8c8d; text-align: center; margin-top: 10px;">
-                    📁 {segment_info}
+                    📁 {file_info}
                 </div>
             </div>
             """
@@ -228,10 +291,17 @@ def process_audio_for_matching(audio_file):
             <div style="text-align: center; margin-top: 25px; padding: 15px; background: #e8f5e8; border-radius: 10px;">
                 <p style="color: #27ae60; margin: 0; font-size: 0.95em;">
                     💡 <strong>How to read results:</strong> The times show where similar segments were found. 
-                    Higher ranked results have better similarity scores.
+                    Use the audio player to listen from the matched timestamp.
                 </p>
             </div>
         </div>
+        
+        <script>
+        function seekToTime(audio, time) {{
+            audio.currentTime = time;
+            audio.play();
+        }}
+        </script>
         """
         
         return formatted_result
@@ -347,11 +417,11 @@ demo = gr.Interface(
     description="""
     <div style="text-align: center; font-size: 1.1em; color: #555; margin: 25px 0; line-height: 1.6;">
         <p><strong>🎯 Upload any audio clip and find similar segments in our music database!</strong></p>
-        <p>Our AI analyzes your audio and finds the most similar segments from known songs.</p>
+        <p>Our AI analyzes your audio and finds the most similar segments from known songs with precise timestamps.</p>
         <p style="font-size: 0.95em; color: #777; margin-top: 15px;">
             📁 Supported formats: MP3, WAV, M4A, FLAC<br>
             ⏱️ Processing time: ~15-30 seconds per file<br>
-            🎼 Database: covers80 collection with segmented analysis
+            🎼 Database: 160 songs with timestamp-based matching
         </p>
     </div>
     """,

inference.py

@@ -9,44 +9,6 @@ def inference(audio_path):
     final_result = result_formatting(result)
     return final_result
 
-def find_closest_segment_file(song_title, target_time):
-    """
-    covers80 폴더에서 해당 곡의 가장 가까운 시간대 세그먼트 파일을 찾음
-    """
-    # 가능한 패턴들로 검색
-    patterns = [
-        f"covers80/{song_title}_segments/*.wav",
-        f"covers80/*{song_title}*_segments/*.wav",
-        f"covers80/{song_title}*/*.wav"
-    ]
-    
-    segment_files = []
-    for pattern in patterns:
-        segment_files.extend(glob.glob(pattern))
-    
-    if not segment_files:
-        return None
-    
-    # 파일명에서 시간 추출하고 target_time과 가장 가까운 것 찾기
-    closest_file = None
-    min_diff = float('inf')
-    
-    for file_path in segment_files:
-        filename = os.path.basename(file_path)
-        try:
-            # 파일명에서 시간 추출 (예: "53.333.wav" -> 53.333)
-            time_str = filename.replace('.wav', '')
-            file_time = float(time_str)
-            diff = abs(file_time - target_time)
-            
-            if diff < min_diff:
-                min_diff = diff
-                closest_file = file_path
-        except ValueError:
-            continue
-    
-    return closest_file
-
 def result_formatting(result):
     """
     get_one_result에서 나온 결과를 포맷팅
@@ -79,14 +41,10 @@ def result_formatting(result):
         # 테스트 레이블에서 정보 추출
         test_time = test_label.get('time', 0) if test_label else 0  # 입력 곡의 시간
         
-        # 가장 가까운 세그먼트 파일 찾기
-        segment_file = find_closest_segment_file(song_title, library_time)
-        
         match_info = {
             'rank': i + 1,
             'score': float(score),
             'song_title': song_title,
-            'segment_file': segment_file,
             'test_time': float(test_time),  # 입력 곡에서 매치된 시간
             'library_time': float(library_time),  # 라이브러리 곡에서 매치된 시간
             'confidence': f"{score * 100:.1f}%",
@@ -108,5 +66,4 @@ if __name__ == "__main__":
         print(f"Rank {match['rank']}: {match['song_title']}")
         print(f"  Score: {match['confidence']}")
         print(f"  Time Match: {match['time_match']}")
-        print(f"  Segment File: {match['segment_file']}")
         print()