Face_recognition: Best way to store the face features in a database
Hello,
A few months ago, I saw the tutorial that, indirectly, originated this library, and, today, I finally had the need to implement a solution that uses facial recognition. However, I got stuck with a big problem when I started planning the data model: what's the best way to store the face features in a database (in order to optimize the search)?
Given the fact that the matches are not based on "perfect matches" (it deals with "similarity"), I'm having a real hard time to build the model. With what I have right now, I would have to load to the memory all the features' sets I have and then execute the comparisons. Given the current application that my project has, that approach wouldn't be a huge problem but... it could become one in the future.
If possible, could you guys, please, give some hints or point some articles/papers on that matter?
GuilhermeCaeiro
>All comments
You could store the face encodings for each user as one database row per user with 128 columns in each row.
Then you could do the face_distance (euclidian distance) calculation yourself as a sql query calculation and sort the database results in order of smallest face distance.
There are a lot of other ways you could do it, too. But that's one idea.
ageitgey
on 6 Apr 2017
Related issues
dickyj
路
3Comments
safaad
路
4Comments
dharam1890
路
3Comments
MLDSBigGuy
路
6Comments
Nijinsha
路
5Comments
Most helpful comment
You could store the face encodings for each user as one database row per user with 128 columns in each row.
Then you could do the face_distance (euclidian distance) calculation yourself as a sql query calculation and sort the database results in order of smallest face distance.
There are a lot of other ways you could do it, too. But that's one idea.