edits

1 files changed, 46 insertions, 0 deletions

diff --git a/check/commands/phash/dedupe.py b/check/commands/phash/dedupe.py
new file mode 100644
index 0000000..3cf60d4
--- /dev/null
+++ b/check/commands/phash/dedupe.py
@@ -0,0 +1,46 @@
+"""
+Dedupe a folder of images
+"""
+
+import click
+import os
+import glob
+
+from PIL import Image
+
+from app.utils.im_utils import compute_phash
+
+@click.command()
+@click.option('-i', '--input', 'opt_input_glob',
+  required=True,
+  help="Input glob to search -- e.g. '../docs/images/*.jpg'")
+@click.option('-t', '--threshold', 'opt_threshold',
+  required=True,
+  default=6,
+  type=int,
+  help="Threshold for PHash hamming distance comparison (0-64, default=6)")
+@click.pass_context
+def cli(ctx, opt_input_glob, opt_threshold):
+  """
+  Dedupe a folder of images
+  """
+  seen = []
+  total = 0
+  for fn in sorted(glob.iglob(os.path.expanduser(opt_input_glob))):
+    total += 1
+    im = Image.open(fn).convert('RGB')
+    phash = compute_phash(im)
+    if is_phash_new(fn, phash, seen, opt_threshold):
+      seen.append({
+        'phash': phash,
+        'fn': fn,
+      })
+  print("checked {} files, found {} unique".format(total, len(seen)))
+
+def is_phash_new(fn, phash, seen, opt_threshold):
+  for item in seen:
+    diff = item['phash'] - phash
+    if diff < opt_threshold:
+      print("{} === {} (diff: {})".format(fn, item['fn'], diff))
+      return False
+  return True