Python machine learning to determine the appropriate me girls?

Hi, just rolling in machine learning and there are questions about in what direction to dig.
Goal: to Train a network on the determination of an appropriate me girls.
Dataset I have to say I'm going to parse something from the Internet and one picture tossed in to the network. If I understand correctly, what weight do I have to put manually each time for each photo?
I want to eventually be able to fill the picture in the script and he gave me a percentage of compatibility.
What algorithms to look at? Where to dig?
Thank you.
March 19th 20 at 08:24
2 answers
March 19th 20 at 08:26
And where does "suitable for your girl"?
In this case, you are planning to train a network on the "photos I like".
One and the same girl can get and not to get into "appropriate", depending on facial expressions, makeup, CLO and so on.
I doubt that the neural network is able to determine the "beautiful or not".
Well, the neural network can identify common features, why not to train to identify all these features in one photograph? - lurline.Kreiger commented on March 19th 20 at 08:29
@lurline.Kreiger, a neural network does not understand human "beautiful". She needs to explain the figures. - Hazle51 commented on March 19th 20 at 08:32
@lurline.Kreiger, If you will be able to collect a sample of girls to a photo of you in the same position, with the same lighting, in General, to make it clear what to compare, maybe.

But again - to recognize this girl and know she's beautiful - are not comparable to the complexity of the task.
Even people don't do it. - Braulio commented on March 19th 20 at 08:35
Okay. Then what did he do?
https://www.youtube.com/watch?v=GAsbZYZk17M - lurline.Kreiger commented on March 19th 20 at 08:38
@lurline.Kreiger, Fun presentation, and trained to work with neural network.

The neural network used in the first place to basically find a picture of the person in the photo, and calculate the proportions to the eyes, nose, mouth, actually face and to obtain these values.

And then he uses separate calculations for trying to understand proportions.
And here the neural network as far as.
To work product then there ogogo - the number of errors due to poor quality pictures or movements exceeds 50%.

Here to do similar for the automatic determination of a person of a certain race, or at least to determine the main genotype - Caucasian, Asian, Negroid, Indian - I think the accuracy could reach the real product. - Braulio commented on March 19th 20 at 08:41
March 19th 20 at 08:28
Don't torture yourself. Download tinder and Bud. There are already all these algorithms and you will only be pushing a like-dislike.
On tinder there are a selection algorithm? Is there not stupidly random geo throws? - lurline.Kreiger commented on March 19th 20 at 08:31

Find more questions by tags Machine learningNeural networksPython