Building a User Maintainable Visualization Dashboard for...
summary
LSST generates 15 Terabytes per night from a 3.2 gigapixel camera and the team builds a deployable multi‑page visualization dashboard so pipeline faults are diagnosed without custom JavaScript. It uses Panel and HoloViews with datashader to rasterize massive point clouds, links interactive plots, and accepts arbitrary pandas filter expressions. Source code and sample datasets are released, the demo handles roughly half terabyte to a terabyte responsively, and some scaling and stability fixes remain.
AI-written summary · about
chapters (10)
- 0:00LSST visualization project intro
- 3:11Static plots fail for pipeline eval
- 7:06Scaling problem in visualization
- 9:13Open source stack for dashboards
- 11:42Dashboard prototype demo
- 17:19Deploying dashboard as web app
- 23:30Advanced filtering and queries
- 25:21Open source vs custom apps
- 26:46Open source community collab
- 27:29Resources and next steps
related talks
40mData-centric exploration using intake, dask, hvplot, and more
36mMaking interactive visualizations easy, inside and outside of notebooks
03h 07mData of Unusual Size: Interactive Visualization
29mHoloViz: Visualization and Interactive Dashboards in Python
37mVisualizing Large Datasets with JavaScript Using deck.gl
40mVisualizing Large Datasets with JavaScript Using Deck.gl
30mSeeing the needle AND the haystack: single-datapoint selection for billion-poin...
46mExplorations in Interactive Visual Analytics