Tweetovi

Blokirali ste korisnika/cu @WePublicHealth

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @WePublicHealth

  1. Prikvačeni tweet
    2. velj

    This week is tweeting on . Warm thanks to for taking us on a US road trip over the past fortnight.

    Poništi
  2. So that's the summary of our unit's latest research! We hope to apply these methods in new ways in the coming year. You can pick up a copy of the article for free here . Thanks to all our co-authors for making it happen!

    Prikaži ovu nit
    Poništi
  3. Looking back at the network chart again, we can see that road injury patterns tend to follow urban design patterns. Even across city types, the effects bleed into one another.

    Prikaži ovu nit
    Poništi
  4. prije 1 sat

    Turns out, more railed transit and more smaller, tighter roads = less injury. Accounting for nearly 90% of variance between cities...

    Prikaži ovu nit
    Poništi
  5. prije 1 sat

    But what features of the maps were related to road trauma and injury rates? To find out, we (or ) counted the colour of every pixel in each map. More black pixels = more roads More orange pixels = more railed public transport

    Prikaži ovu nit
    Poništi
  6. prije 1 sat

    So how are these city types related to road truama? Well it turns out, there's quite a difference in performance between them across modes as you can see here - and this is controlling for within-city economic activity

    Prikaži ovu nit
    Poništi
  7. prije 1 sat

    And this is where they are located across the globe

    Prikaži ovu nit
    Poništi
  8. prije 1 sat

    You can see the 9 different types and some examples of their maps, below. Where do you think your city fits? How about others you may have visited? Which ones look like ?

    Prikaži ovu nit
    Poništi
  9. prije 1 sat

    Eventually after millions of images, predictions and confusions, this graph built up, featuring every city in the dataset and their relationship to one another. Using some fancy stats, we were then able to classify them into 9 major 'modules' or city types.

    Prikaži ovu nit
    Poništi
  10. prije 1 sat

    Every time the computer confused two cities from each other, it linked them to one another in spatially representative graph, kind of like this. You can see how the images on the left are kind of similar, but the ones on the far right are quite different.

    Prikaži ovu nit
    Poništi
  11. prije 2 sata

    Now most models like this are interested in getting the model really accurate - but we were interested in when the model got it wrong. Why? Because like our Tokyo example, when the model got confused between two cities, it showed the cities had similar urban design features...

    Prikaži ovu nit
    Poništi
  12. prije 2 sata

    So - we had a computer built that could do it! We called it Hal. We collected 1.7 million map images from 1692 cities around the world - 1000 neighbourhood-sized maps from each place. We fed the images to the computer until it got good at recongising city maps - 86% good

    Prikaži ovu nit
    Poništi
  13. prije 2 sata

    So for example - what makes this Melbourne, and that Tokyo?

    Prikaži ovu nit
    Poništi
  14. prije 2 sata

    That was a lot harder because its not the way we usually see things. But it raised a question. Could we train a computer to recognise cities from one another simply by showing it thousands of small maps - just like training on pics of cats vs dogs?

    Prikaži ovu nit
    Poništi
  15. prije 2 sata

    That's right - they are the same cities as before...

    Prikaži ovu nit
    Poništi
  16. prije 2 sata

    OK - Hold that thought. If you did confuse any of the cities, why? What features of the image made you think it was from somewhere else? Now - Here's a second chance. Can you guess where these 6 cities are from?

    Prikaži ovu nit
    Poništi
  17. prije 2 sata

    No cheating, now.... Answers... 1. 2. 3. 4. 5. 6. How many did you get right? Let's be honest, did you confuse Tokyo for Paris?

    Prikaži ovu nit
    Poništi
  18. prije 2 sata

    So we wanted to investigate whether city and urban design types affect road trauma. Here's a pop quiz for you global types - Can you quickly name these cities?

    Prikaži ovu nit
    Poništi
  19. prije 2 sata

    In particullar - road trauma is directly related to the way in which we set up our cities. The vehicles we use, the mix of vehicles, the public transport, the speeds, our behaviour all contribute. And road trauma takes ~1.3 million lives each year & injures 50+ million.

    Prikaži ovu nit
    Poništi
  20. prije 2 sata

    But it's an important question and topic because Urban design and cities can affect all these things, below, which are the diseases and illness that kill most of us across the ages.

    Prikaži ovu nit
    Poništi
  21. prije 2 sata

    A common question in among urban design and health researchers is, 'What does healthy city design look like?". And to be honest, it's not exactly an easy question to answer.

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·