How a computer seeks, interprets internet dance information and represents them as an artistic form, a form you can just get, not process.
DATE: May 2016
DURATION: 1 month
TOOLKITS: python, django, D3.js
COLLABORATORS: Yuheng Zhan
SPONSORSHIP & EXHIBITIONS: ITP SpringShow 2016, Google XStory Annual Showcase 2016, NYC Media Lab Annual Submit 2016.
restage wants to disrupt our perspective of finding and looking at dance. It is a data visualization and generative media creator powered by the Internet dance metadata that are produced by a web crawler. restage generates meta media for dance using web crawling, natural language processing and etc. Then represents these meta media by remixing them into different artistic forms.
ReStage seeks to use storytelling and look for different artistic ways to tie the internet dance information together into new art forms. In order to open a new window for people to understand dance, and offer further inspirations. Current ways include photo mosaic creation, mosaic gif, and style transfer.
USER EXPERIENCE & COMPONENTS EXPLANATION:
ReStage contains several main parts: a dance data crawler, web search engine, visualization of text based information for each dance and a dance art collage generator.
For each dance the information contains title, relevant artists name and links, NYT review links, other links, performance images, videos, and especially keywords and featured keywords. User could select featured keywords to generate a photo mosaic for the dance and then further see a generate gif. The result wants to assist the physical experience of seeing a dance performance, use elements that is not relevant to dance to describe a dance and help to open another window for people to interpret them.
Dance web search engine:
In order to collect Internet dance data, restage has a database/search engine specifically for New York dance performances. It starts from crawling the New York Times dance subject articles after 2007, extracting basic performance information(performance title, artists names, outlinks, keywords and so on) based on some typical dance review writing patterns and natural language processing. Then uses these information and further expands to collect other information as well as media include images and footages.
Users can start from searching a dance performance by either the piece title or the artist names, like a famous choreographer or dancer. Then they will see the results.
By clicking the blocks, users are further guided to see the details of each piece.
Dance art collage generator:
ReStage tries to answer the basic questions from the data collection part: What the piece was about and what is the choreographer trying to do? What movement vocabulary or style used? How did the audience feel? And what is the inspiration? ReStage uses some language processing methods to deconstruct these descriptive text information(review, description) into keywords/featured keywords. Strong objective keywords are further returned with featured images. ReStage then uses these 3 main source information/meta media: video, image, and text to automatically generate different art collages for each dance.
Users are encouraged to select interested featured keywords that describe the dance, a performance image, and a threshold that indicates the number of divisions. Then the performance image is recompiled into a mosaic image based on pictures that are generated from a keyword image search.
restage has another featured image database. After the user selected featured keywords, if there was no featured image list in the database for a keyword, a google image search is called and update the database, then return a list of featured images responding to the selected featured keywords with RGB value. Images are decomposed into smaller segments that are made from featured images returned and replacing sections of the base image with a similar overall color.
Users can also play with different level of resolution and put images together into a gif :
To see more generative images, this post would offer you a better understanding: http://www.streamgao.com/?p=2336