.gitignore

@@ -102,4 +102,4 @@ ENV/
 *.swp
 *.swo
 
-mcdata/
+test/

main.py

@@ -1,18 +1,18 @@
 import glob
 import os
 import hashlib
-import time
-import traceback
-from multiprocessing import Pool
-import queue
 from PIL import Image, UnidentifiedImageError
 
+import acoustid
+import chromaprint
 from imagehash import average_hash
 from videohash import VideoHash
 from videohash.exceptions import FFmpegFailedToExtractFrames
 
 hashes = set()
+audio_hashes = []
 delete = set()
+hash_lookup = {}
 
 filenames = [x for x in glob.glob('**', recursive=True) if os.path.isfile(x)]
 
@@ -21,11 +21,11 @@ print('ENTER to continue, ctrl+c to cancel.')
 try:
     input()
 except KeyboardInterrupt:
-    print('Cancelled.')
+    print('\nCancelled.')
     os._exit(0)
 
-print('Sorting file list by smallest size...')
-filenames.sort(key=os.path.getsize, reverse=False)
+print('Sorting file list by size...')
+filenames.sort(key=os.path.getsize, reverse=True)
 
 print('Deduplicating by md5 hash...')
 
@@ -39,82 +39,87 @@ for filename in filenames:
         hasher.update(buf)
         digest = hasher.hexdigest()
         if digest in hashes:
-            print('Found digest', digest, 'collision for', filename)
             delete.add(filename)
         else:
             hashes.add(digest)
 
 print('Found', len(delete), 'duplicates by md5 hash.')
-
-time.sleep(4)
 print('Deduplicating by media fingerprint...')
 
 def get_image_hash(filename):
-    basename = os.path.basename(os.path.dirname(filename))
     try:
         image = Image.open(filename)
-        return basename + str(average_hash(image)) + '_image'
+        return str(average_hash(image)) + '_image'
     except UnidentifiedImageError:
         return None
 
-def get_video_hash(filename):
-    basename = os.path.basename(os.path.dirname(filename))
+def get_audio_hash(filename):
     try:
-        v = VideoHash(path=filename)
-        digest = str(v.hash_hex)
-        v.delete_storage_path()
+        fp = acoustid.fingerprint_file(filename)[1]
+        values, _ = chromaprint.decode_fingerprint(fp)
+        return chromaprint.hash_fingerprint(values)
+    except acoustid.FingerprintGenerationError:
+        return None
 
-        return basename + digest + '_video'
+def get_video_hash(filename):
+    try:
+        return str(VideoHash(path=filename).hash_hex) + '_video'
     except FFmpegFailedToExtractFrames:
         return None
 
 count = 0
 total = len(filenames)
 
-def hasher(filename):
-    if filename in delete: return None
+for filename in filenames:
+    count += 1
+    print('Hashing file', count, '/', total, end='\r')
 
-    print('Hashing file:', filename)
+    if filename in delete: continue
 
     try:
-        digest = get_image_hash(filename)#  or get_video_hash(filename)
+        digest = get_image_hash(filename) or get_audio_hash(filename) or get_video_hash(filename)
     except KeyboardInterrupt:
-        print('Cancelled.')
-        os._exit(0)
+        print('Skipping media hashing.')
+        break
     except BaseException as e:
         print()
         print('Exception', e.__class__.__name__, str(e), 'while hashing:')
         print(filename)
-        print(traceback.format_exc())
-        return None
+        continue
 
-    if not digest: return None
+    if not digest: continue
 
-    return (filename, digest)
-
-with Pool() as pool:
-    results = pool.map(hasher, filenames)
-
-print('Finished hashing.')
-print()
-print('Checking digests:')
-print()
-
-for result in results:
-    if not result: continue
-    print(result)
-
-    filename, digest = result
+    hash_lookup[digest] = filename
 
+    if type(digest) == int:
+        for h in audio_hashes:
+            if bin(digest ^ h).count('1') <= 5:   # TODO adjust?
+                delete.add(filename)
+                print()
+                print(digest, filename, 'close to', h, hash_lookup[h])
+                break
+        else:  # for
+            audio_hashes.append(digest)
+    else:
         if digest in hashes:
-        print('Found digest', digest, 'collision for', filename)
             delete.add(filename)
         else:
             hashes.add(digest)
 
 print()
-print('Found', len(delete), 'total duplicate files.')
+print()
+for dupe in delete:
+    print(dupe)
 
-for dupe in sorted(list(delete)):
-    print('Deleting:', dupe)
+print()
+print('Found', len(delete), 'total duplicate files. Delete them?')
+print('ENTER to continue, ctrl+c to cancel.')
+try:
+    input()
+except KeyboardInterrupt:
+    print('\nCancelled.')
+    os._exit(0)
+
+print('Deleting...')
+for dupe in delete:
     os.remove(dupe)

requirements.txt

@@ -1,13 +1,19 @@
-Brotli==1.1.0
-certifi==2023.7.22
+audioread==3.0.0
+Brotli==1.0.9
+certifi==2022.12.7
+charset-normalizer==3.1.0
+idna==3.4
 imagedominantcolor==1.0.1
 ImageHash==4.3.1
-mutagen==1.47.0
-numpy==1.26.1
-Pillow==10.1.0
-pycryptodomex==3.19.0
+mutagen==1.46.0
+numpy==1.24.2
+Pillow==9.5.0
+pyacoustid==1.2.2
+pycryptodomex==3.17
 PyWavelets==1.4.1
-scipy==1.11.3
+requests==2.28.2
+scipy==1.10.1
+urllib3==1.26.15
 videohash==3.0.1
-websockets==11.0.3
-yt-dlp==2023.10.13
+websockets==11.0
+yt-dlp==2023.3.4