I have the following code that allows me to find images of equal has (identical), but say I wanted to just find images with a hamming distance under a certain number, can that be incorporated into django querysets, or raw sql somehow? I don't want to fetch everything and compare with python because that's very very slow and I many many images.
Current code:
def duplicates(request):
duplicate_images = []
images = Image.objects.all()
for image in images:
duplicates = Image.objects.filter(hash=image.hash).exclude(pk=image.pk)
for duplicate in duplicates:
duplicate_images.append([image, duplicate])
if len(duplicate_images) > 1000:
break
Here is how to implement this using a postgres extension:
https://github.com/eulerto/pg_similarity
Installation:
$ git clone https://github.com/eulerto/pg_similarity.git
$ cd pg_similarity
$ USE_PGXS=1 make
$ USE_PGXS=1 make install
$ psql mydb
psql (9.3.5)
Type "help" for help.
mydb=# CREATE EXTENSION pg_similarity;
CREATE EXTENSION
No you can make a django queryset with a custom "WHERE" clause in order to use the hamming_text function
image = Image.objects.get(pk=1252) # the image you want to compare to
similar = Image.objects.extra(where=['hamming_text(hash,%s)>=0.88'],
params=[image.hash])
and voila, it works!
note: the hamming distance here is automatically normalized so 0 means completely different and 1 means identical.