Awesome

University-Logo-Deep-Learning-Community-Detection

About ourself

Team members:

Langford-tang , 2016, Southern University of Science and Technology
HUSTERGS ,2017, Huazhong University of Science and Technology
JacinthGDRGN , 2017, East China Normal University
zmw1216 , 2017, Xi'an Jiaotong University

Team leader: Langford-tang

About the project

Basically we crawled most university logos from all over the world including 14 countries or cities, and we feed the dataset into autoencoder neural network to get the high-dimension representation of the logos, so we can calculate the similarity among them and get the distance matrix. And we firstly apply k-means and do community detection within every cluster, hoping to find something interesting and we actually did!

part of our interesting findings

for more detailed information, please look into the poster and powerpoint we made, If you like this project, you can give us a :star:

Front end

cd Frontend
python -m http.server

open http://0.0.0.0:8000 in your browser

Frontend folder is no longer maintained, please go to the Backend folder which integrate the front end

Back end

cd Backend
sudo python ./server.py
# open 0.0.0.0 in your browser
# you can change the port and host in server.py if prefer not to run with root privilege

still under develop

data

data_200_pixel all the data after manually select, format, crop, resize

Cleaned.v2_format_jpeg.zip all the raw data after manually select and format

src

*.py related to neural network

*.ipynb

University_in_*

crawl data using simulation browsing and Google Image Search Engine
SelectData.ipynb

Select data from the whole dataset to do algorithms comparison
China-Sample-3d.ipynb

Apply community detection algorithms to the distance matrix generated from neural network, and produce the json file to feed into the 3d-force-graph framework
mosaic.ipynb

this is just a by-product

generate mosaic picture with university logos we crawled using photomosaic python package

from

to