0

Im using the following https://github.com/ageitgey/face_recognition to process images with faces in. Ive used it a few times, but im wanting to get a bit more serious with it, and store a large amount of images and be able to query across them all quickly. So the obvious would to store the 128 Vector in a database. At the moment im wanting to use MySQL as its more in my comfort zone while im learning.

From what i have read i would need to create a column for each of the 128 vectors, and use FLOAT. But i'm cant find anything more on specifics like FLOAT(10,6) etc, or if i need to create any specifics regarding indexes.

The next thing would be then be the Query to search the known vectors with the vectors taken from an image im wishing to search. But i cant find any actual working examples of this.

Neil
  • 95
  • 1
  • 1
  • 7
  • There is no `FLOAT(10,6)` but `DECIMAL(10,6)`. – Paul Spiegel Jan 27 '19 at 17:34
  • A while back when storing Long, Lat i read to use FLOAT(10,6) when defining my columns. When googling now i get several posts coming back with FLOAT(10,6) eg https://stackoverflow.com/questions/12504208/what-mysql-data-type-should-be-used-for-latitude-longitude-with-8-decimal-places – Neil Jan 27 '19 at 17:40
  • OK sorry, "MySQL permits a nonstandard syntax: FLOAT(M,D)". But you probably want to use `DECIMAL` instead. Read the answer from the link in your comment. – Paul Spiegel Jan 27 '19 at 17:46
  • Its just a mathematical formula. But applying this on every row with thousands of images will slow down the performance. Try exploring in memory option. Database can act as a backup where you can restore the encodings from in case of a server restart like situation. – Manish Mudgal Jun 20 '19 at 10:37

0 Answers0