https://github.com/hadro/pix-plot-dpla
PixPlot remix with images from DPLA
https://github.com/hadro/pix-plot-dpla
Last synced: over 1 year ago
JSON representation
PixPlot remix with images from DPLA
- Host: GitHub
- URL: https://github.com/hadro/pix-plot-dpla
- Owner: hadro
- License: mit
- Created: 2018-06-29T20:00:40.000Z (about 8 years ago)
- Default Branch: master
- Last Pushed: 2018-07-04T15:23:06.000Z (about 8 years ago)
- Last Synced: 2025-01-18T17:24:05.143Z (over 1 year ago)
- Language: JavaScript
- Homepage: https://hadro.github.io/pix-plot-dpla/
- Size: 197 MB
- Stars: 3
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: license.md
Awesome Lists containing this project
README
# PixPlot
Remix of PixPlot viz based on Yale DH Lab's original [PixPlot repository](https://github.com/yaledhlab/pix-plot).
See remix viz here: https://hadro.github.io/pix-plot-dpla/
## Intro text from site
This page visualizes 7,384 "Fourth of July" (and related search) images from the Digital Public Library of America (DPLA), which aggregates cultural heritage metadata from thousands of institutions.
Under the hood is a WebGL model of how the Inception Convolutional Neural Network perceives "relatedness" and then a mapping of the computed relationships into 3-dimensional space using UMAP algorithm such that similar images appear near each other.
Except for a few small modifications, this is based on the work and code published by the Yale Digital Humanities Labs in their PixPlot repository, which I learned about through an excellent video presentation from a Coalition for Networked Information (CNI) project briefing.
---
Controls:
Click the clusters on the left to zoom to those areas, or use the mouse and/or keyboard to browse on your own. Scroll to zoom in, or use the up/down keys. Right click to rotate.
Clicking on an item will center the camera to that image, and double-clicking an item will take you to the record for that image in DPLA.
## Acknowledgements
I couldn't have done this without the awesome code and materials published by the Yale Digital Humanities Lab. See their original repo for much more thorough background and acknowledgements in terms of the data science and techniques that underlie this visualization.