We've long wanted to be able to support multiple streetscape datasets (e.g., from Mapillary, Apple Maps, Bing Maps).
I spent a bit of time today (while multitasking on Zoom calls) exploring Mapillary.
Mapillary has many advantages—it relies on crowdsourced streetscape data; however, this data is far sparser, more diverse in quality, and not full 360.
Here's the current coverage map of Mapillary:
Here's the coverage map of Seattle. Looks good on a quick scan but you quickly see many streets without coverage (e.g., in Wallingford, Queen Anne, Ballard, Ravenna, etc.):
It's worse in less populated areas. Here's Bremerton, for example:
Here's Poulsbo:
Here's a comparison with GSV:
People contribute a variety of data from dashcams, mobile phone images, etc. Here's an example from a dashcam in Seattle but doesn't let you pan the image (so it's a fixed view pointing straight ahead)
So, my sense is that Mapillary is still not a good enough dataset for us.
On the other hand, it could be a possibility if people are uploading their own 360 images from on-the-ground audits.
Here's an interesting analysis of Apple's drivethroughs: https://www.justinobeirne.com/look-around-rollout. Some cool tidbits in here such as Apple's Look Around covering every metropolitan area in Canada (35 in total) + thousands of rural towns and villages between them. Look Around is also now available in all US counties
We've long wanted to be able to support multiple streetscape datasets (e.g., from Mapillary, Apple Maps, Bing Maps).
I spent a bit of time today (while multitasking on Zoom calls) exploring Mapillary.
Mapillary has many advantages—it relies on crowdsourced streetscape data; however, this data is far sparser, more diverse in quality, and not full 360.
Here's the current coverage map of Mapillary:![image](https://github.com/user-attachments/assets/e12590a9-3caf-4757-b26e-8543136d9ddf)
Here's the coverage map of Seattle. Looks good on a quick scan but you quickly see many streets without coverage (e.g., in Wallingford, Queen Anne, Ballard, Ravenna, etc.):![image](https://github.com/user-attachments/assets/cfdf1816-b785-48e9-9f13-c650ccd236ab)
It's worse in less populated areas. Here's Bremerton, for example:![image](https://github.com/user-attachments/assets/327c84df-8415-43da-80d3-e0b73a00df00)
Here's Poulsbo:![image](https://github.com/user-attachments/assets/ef09b4bc-3a68-4c16-95a2-b90d619310a8)
Here's a comparison with GSV:![image](https://github.com/user-attachments/assets/68baa331-ad93-4062-8430-fd1739352218)
People contribute a variety of data from dashcams, mobile phone images, etc. Here's an example from a dashcam in Seattle but doesn't let you pan the image (so it's a fixed view pointing straight ahead)![image](https://github.com/user-attachments/assets/1ecf5c94-b38c-4bdd-87b6-894635705980)
So, my sense is that Mapillary is still not a good enough dataset for us.
On the other hand, it could be a possibility if people are uploading their own 360 images from on-the-ground audits.