- This application is a Matching SNS created based on Instagram UI
- Label the image posted by the user with VGG16.
- Create a user feature vector based on the label.
- Calculate cosine similarity between users from user feature vectors and recommend users with high similarity.
VGG 16 is a 16-layer CNN model learned with a large-scale image dataset called "ImageNet." It was announced in 2014. It is a famous It is one of the learned models
- このアプリケーションは、インスタグラムのUIをベースに作られたマッチング系SNSです。
- ユーザーが投稿した画像に対して、VGG16でラベル付けをします。
- そのラベルを元にユーザー特徴ベクトルを作成します。
- ユーザー特徴ベクトルからユーザー間のコサイン類似度を計算し、類似度が高いユーザーをレコメンドします。
VGG16というのは、「ImageNet」と呼ばれる大規模画像データセットで学習された、16層からなるCNNモデルのことです。2014年に発表されました。
様々な研究で利用されている，有名な学習済みモデルの1つです。
The information posted by the user is listed as a timeline.
Imagenet inference is Inference result of VGG16
Probability is Probability of the above inference
You can easily post images from the post form
The most important feature is user base filtering from the user's posted image and recommends similar users
Also view their posts on the user page
$ mkdir <YOUR DIR> && cd <YOUR DIR>
$ python -m venv venv
$ source venv/bin/activate
$ git clone [email protected]:10380r/Mimic-Instagram.git
$ cd Mimic-Instagram
$ pip install -r requirements.txt
$ ./manage.py migrate
$ ./manage.py createsuperuser
$ ./manage.py runserver
login at admin page
- create users
Notes: When entering User information, change the permission to Staff status. Also select Choose all for User permission.
Then, you can try sns. at https://localhost:8000/app
See reqirements.txt for details
python 3.7.2
django 2.1.5
Mac OS Mojave 10.14.3 (I am trying to execute it only on this OS)
VGG16
vue.js network You can change the drawing style of the network by changing vue.
- Follow function
- DirectMessages function
- Enable to post videos