User:Ruben/Prototyping/News images
On Digital Interpretation
This version is a continuation on the SIFT project.
It is a merge with the photobook project, which at that time was about the 'forever new' of the digital (as this is probably a major contrasting properties of the digital book vs. the physical book).
It consists of several parts:
The crawler loops over several rss feeds of newswebsites and collects their content and images. When a new item is found, it saves it to the database and it notifies the 'server'.
The server is the central hub of the project. It serves an html file out on port 8888, which in turn starts a socket-connection to the same server. When it receives notifications of new newsitems from the crawler, it sends this update to the connected browsers, so they in turn switch to that image. Simultaneously the server processes the image using SIFT to find its features (and draw these on a copy of the image). Then it starts a second python process which also connects to the server using a socket connection. This process takes the features of the new image and runs them over the existing images on disk. When a match is found, it's corresponding news item is collected from the database and then send to the server and all connected browsers. This matching script is interrupted by the server when a new newsitem comes in.
The browser has a d3.js treemap running [1], which is updated to add each new match. The size of the items is based on the number of features that matches. And the styling of the items is based on their size (small ones having no description and images etc.)
For non-technical considerations/intentions and continuation plans see Thematic Project 2.
Full source code and description is available at BitBucket.
- ↑ I first tried to adapt several other scripts like shuffle.js and masonry, but these all lead to nothing.