Uploads from stevefaeembra, tagged visualisation

colour analysis of google images

nobody@flickr.com (stevefaeembra) — Sat, 09 Mar 2013 14:16:29 -0800

doing an analysis of the most common colours in certain google image searches.

Worked out the 15 most common colours for each of a number of Google image searches, using k-means analysis

The barcodes show how common each colour is; the wider the stripe, the more common that colour is.

Generated using a python script, knocked into shape with InkScape.

International Land Deals

nobody@flickr.com (stevefaeembra) — Sun, 29 Apr 2012 04:10:30 -0700

stevefaeembra posted a photo:

see notes below (flickr descriptions a bit broken at the moment)

osm-analysis-using-gephi

nobody@flickr.com (stevefaeembra) — Tue, 06 Dec 2011 14:07:23 -0800

stevefaeembra posted a photo:

using Gephi open source graph analysis software to analyse road networks from OpenstreetMap xml data

used a python script to convert an osm xml file to a gexf network file, then used gephi to do network analysis. OSM nodes map to network nodes, and road segments map to edges. Used the geolayout plugin to make sure the nodes stay put and don't rearrange themselves into a hairball :)

In this case, the more orange the road, the more important (and/or long and straight) it is, so the more traffic it's likely to get.

Font fingerprints

nobody@flickr.com (stevefaeembra) — Fri, 07 Oct 2011 14:09:55 -0700

stevefaeembra posted a photo:

Visualising several common fonts, to look for common patterns.

Using the Processing language, superimposed letters of the alphabet (in proportion to how often they appear in English text), using different fonts.

Blue shows where less ink was used, white shows where more ink was used.

Notes: Assumed capital letters were less frequent than the lower case equivalents by a factor of 10. Ignored punctuation marks and digits.

Edinburgh Road Network analysis

nobody@flickr.com (stevefaeembra) — Sat, 17 Sep 2011 13:21:20 -0700

stevefaeembra posted a photo:

Using OSM data to analyse road networks; this is a map of Edinburgh.

The brighter the road, the more places you can reach on foot (or by car) by walking or driving up to 1km. Darker roads are more remote, more 'out of the way'. Dark 'holes' represent parks, golf courses and so on.

Mouseover for annotated notes.

map created using data from OpenStreetMap

Tech notes / Caveats :
- I only chose to use the locations of the roads, and ignore their importance. Footpaths were omitted.
- Didn't download the whole city, just a few tiles.
- Can be skewed by some areas of detailed mapping, such as Craigleith Retail Park, whose car park acts like a "city within a city" :-)

commuter trails 2.0

nobody@flickr.com (stevefaeembra) — Thu, 11 Aug 2011 13:23:22 -0700

stevefaeembra posted a photo:

graphic comparing the three different routes I've needed to use to get to Glasgow over the last two days (with thanks to the lovely Scottish weather and the "wrong sort of rain on the line" ;-)

Width of line is in proportion to speed; the big circles are the approximate edges of the two cities (bit of artistic licence here)

rail journey from Winchburgh to Linlithgow

nobody@flickr.com (stevefaeembra) — Fri, 10 Jun 2011 12:41:32 -0700

stevefaeembra posted a photo:

a slitscan timeline visualising part of the rail journey from Edinburgh to Linlithgow, looking south.

Video frames were compressed to 1 pixel wide, then stacked horizontally into a timeline using a script written in Processing.

Shows how the view changes over time - leafy cuttings, trees, and an occasional sweeping vista. The black square marks a tunnel.

Chess analysis

nobody@flickr.com (stevefaeembra) — Sun, 27 Mar 2011 12:37:48 -0700

stevefaeembra posted a photo:

An analysis of several thousand chess games, trying to visualize patterns in the moves made.

Each type of piece has its own heatmap.

Each square shows how often that type of piece (for either player) moved into that square. Red means lots of times, blue means fewer times. The boards follow the traditional layout convention (white at the bottom, black at the top).

The area of the circles shows how often those moves resulted in a capture (or in the bottom diagram, a check). The same scale is used throughout, so you can compare how dangerous each type of piece is in different parts of the board.

Python script to parse the PGN files, Processing to do the rendering.

prague photographic hotspots

nobody@flickr.com (stevefaeembra) — Sat, 29 Jan 2011 13:40:32 -0800

stevefaeembra posted a photo:

Part of an ongoing project to examine the photographic points of interest, using the Flickr API.

This image shows the most commonly photographed parts of Prague. Red cells are the most geotagged, blue the least geotagged.

Hotspots are the Castle, The Museum, Charles Bridge and the Old Town Square. No major surprises there ;-) Some other hotspots are scattered around the edges of the city; I suspect some of these are conference or concert venues.

Hotspot image was done in Matplotlib using a hexbin plot, and used Google Earth to project as an image overlay onto OpenStreetMap layers.

Data Copyright OpenStreetMap and contributors, CC-BY-SA, data via www.geofabrik.de/

analysis of aspect ratios on Flickr

nobody@flickr.com (stevefaeembra) — Sun, 26 Dec 2010 01:53:36 -0800

stevefaeembra posted a photo:

an experiment using the Flickr API. Press L to see bigger.

These diagrams reveal the usage of different aspect ratios* of images on Flickr with different tags. Faint blue is least common, bright red most common. Portrait orientations on the left, landscape on the right. Square images are counted as being portrait so they appear on the left.

(* ratio of image width to height)

Some observations... not too scientific

Most common ratio seems to be 3:2, followed by 4:3. This suggests more DSLR than compact.. Anything that isn't close to one of these two ratios is probably cropped or stitched (e.g. panoramas).

"Party" shows little variation, tending to stick to the common camera sizes. In other words, people don't tend to crop their party photos, but use them straight from camera.

"HDR" shows even more extreme letterbox compositions than 'Panorama'. It also bucks the trend in having 4:3 as the most common aspect ratio.

In most cases I've tried so far, it seems people prefer to take pictures in landscape orientation. The exception is "square" which (funnily enough) returns almost all square images, and "portrait" :-)

"Portrait" has a lot of variation in the ratios, suggesting these images are getting cropped rather than being used at the original size.

Took first 1000 images found when searching for the given tag, and looked at those had an original size available. Ratios rounded to nearest .01 for binning purposes. Coded in C# with Flickr.net.

eurovision voting patterns visualised

nobody@flickr.com (stevefaeembra) — Sat, 26 May 2012 15:16:17 -0700

stevefaeembra posted a photo:

visualising the voting pattern of eurovision nations in the last 20 years.

countries are placed in their geographic position, with small nudges for aesthetic reasons

the dot size is the number of votes received, and line colours/widths reflect the total number of votes cast between nations.Arcs go clockwise from the giver of votes to the receiver of votes.

Happy 30th birthday, ZX Spectrum!

nobody@flickr.com (stevefaeembra) — Mon, 23 Apr 2012 12:36:04 -0700

stevefaeembra posted a photo:

today is the 30th birthday of the launch of the ZX Spectrum, the classic 8-bit computer.

This visualisation shows which parts of the ROM have been accessed, and how often - a "code coverage" report of sorts.

- Opcodes (instructions) appear in white, and operands/data in red.
- Brighter areas have been accessed more often
- Areas in dark grey were never accessed.
- The ROM is shown in vertical strips of 256 bytes, reading from left to right and top to bottom.

This was produced by customising the wonderful java emulator qaop with some instrumentation code, and plotting the results in Processing.

This was the result of playing the "Horizons" educational software, which should have exercised a lot of the BASIC, graphics and floating point routines.

protanope colour blindness simulation

nobody@flickr.com (stevefaeembra) — Thu, 22 Sep 2011 15:04:15 -0700

stevefaeembra posted a photo:

developing a script to convert images to appear as they would to dichromat colour-blind viewers - in this case, protanopia.

top half of image is original, bottom half as seen by someone with protanopia.

seems to tally well with Gimp's view filters for colour blindness (well, for Protan at least) and it makes the numbers disappear from the Ishihara dot tests.

Based on this paper

OpenstreetMap user participation map for London

nobody@flickr.com (stevefaeembra) — Mon, 27 Jun 2011 14:02:34 -0700

stevefaeembra posted a photo:

Map of London showing the number of distinct users in the most recent changeset on OpenStreetMap. Caveat: only shows a snapshot of the map as it stands, not the historical changes.

Blue areas show a lower number of users, orange a higher number of users. There seems to be a distinct "band" of detail running from North to South, where the map is being built in more detail. The "density" of dots shows how detailed the map is.

I've tried to make this "colour-blind friendly". If you're colour-blind I'd be interested to hear any feedback on how well this works for you :)

Generated using Processing.

Data copyright OpensteetMap and its contributors, CC-BY-SA.

hamlet

nobody@flickr.com (stevefaeembra) — Sun, 29 May 2011 17:46:19 -0700

stevefaeembra posted a photo:

A timeline showing the comings and goings of characters in Shakespeare's "Hamlet".

visualizing how football (soccer) clubs derive their income

nobody@flickr.com (stevefaeembra) — Fri, 11 Feb 2011 07:22:06 -0800

stevefaeembra posted a photo:

data via Guardian Datablog.

Position of each club depends on how the club's revenues are made up : Matchday ("bums on seats"), Commercial (Sponsorship) and Broadcasting rights. Colour of club's dot is in proportion to it's revenue; higher revenue is lighter.

Seem to be three noticeable clusters. German teams are mostly Commercial, Italian clubs mostly Broadcasting, and UK and other European clubs tend to be a more balanced mixture.

Created using Orange and Inkscape.

1 o'clock gun map

nobody@flickr.com (stevefaeembra) — Fri, 14 Jan 2011 10:01:52 -0800

stevefaeembra posted a photo:

today was the 150th anniversary of the One O'Clock Gun at Edinburgh Castle.

This was inspired by the original One o'clock gun map which shows how long the sound takes to reach different parts of the city.

Each circle represents 1 second. It's surprising how slow sound travels when you look at the scale of a city - it takes about 10 seconds for the sound wave to reach Leith.

Map via OpenStreetMap

This is a screencap, there is a real-time animated version here on OpenProcessing (requires Java)

world terrain as a stacked bar chart

nobody@flickr.com (stevefaeembra) — Fri, 07 Jan 2011 14:58:23 -0800

stevefaeembra posted a photo:

Some experimental cartography.

This is the NASA blue marble image, organised as a stacked bar chart :)

The x axis is longitude.

The pixels in each column were sorted vertically by brightness, so that ice and snow drifts to the top, ocean drifts to the bottom, and desert/grassland are in the middle. Antartica has drifted to the top of the screen to join the Arctic and the Tundra.

This helps visualise a number of things;

- the sheer area of ocean (around 70% of the earth's surface)
- how land type varies by longitude

Done using Processing. Blue Marble courtesy of NASA.

partitioning a city using flickr API

nobody@flickr.com (stevefaeembra) — Fri, 17 Dec 2010 05:53:15 -0800

stevefaeembra posted a photo:

This is a map of the centre of Edinburgh, Scotland using data from the FlickR API. The aim was to see if it's possible to use Flickr geotagged data to get a feel for the geography of a city.

Over 100,000 photos were clustered into 20 areas, using k-means analysis. Mouse-over the image for some notes on what parts of town are where.

This particular version matches my 'mental map' of Edinburgh quite well.

It differs from Flickr's shapefile divisions, which use the "which part of town did you take this photo?" geotagging option to deduce the boundaries of each neighbourhood.

Number crunching and rendering using Orange Canvas.

jobseeker's allowance figures october 2010

nobody@flickr.com (stevefaeembra) — Thu, 18 Nov 2010 14:17:55 -0800

stevefaeembra posted a photo:

Infographic showing how the total percentage of Jobseeker's Allowance claimants varies across the UK's constituencies.

Props for the template svg file at Wikimedia Commons here. Munged with data from ONS via the Guardian Datablog. You wouldn't believe how much data cleansing you need with the names of constituencies :)

Disclaimers: Claimant counts are much lower than the actual number out of work - doesn't count those 'economically inactive', those not elegible or on other benefits. Figures not seasonally adjusted.