2D-Packing Images 2014
--Thiebaut 10:34, 8 September 2014 (EDT)
The goal of this project is to create and modify a packing algorithm that takes aesthetics into account in order to conduct a large-scale comparative analysis. First, the heuristic, which is a greedy best-fit heuristic, was first implemented in Python for easy prototyping. The project was then transferred to Java to improve packing speed. In Java, a multi-threaded version was implemented to allow for simultaneous packing. Since the goal is to work with large datasets, the images are stored on a remote server, rather than a local server. As such, a database needed to be created to store information on each image's file location, geometry and statistics. A future goal would be to add the appropriate statistics to each image as currently the statistics used are randomly generated. Future work will involve the progressive decomposition of images, from blurry to crisp, to allow for faster real-time viewing.
Updated as of 12/2/14
- --Thiebaut 16:07, 30 September 2014 (EDT) Interesting graph. It's not the result of a packing algorithm, though, but it's a tree. It's much simpler and faster to generate than our packing.
- --Thiebaut 16:07, 30 September 2014 (EDT) cool dynamic graph. We might be able to use that to repack on the fly and make images of recent edits come up bigger than the others...