With gsvp-dl, an open source solution written in Python, you are able to download millions of panorama images off Google Maps Street View.

Unlike other existing solutions (which fail to address major edge cases), gsvp-dl downloads panoramas in their correct form and size with unmatched accuracy. Using Python Asyncio and Aiohttp, it can handle bulk downloads, scaling to millions of panoramas per day.

It was a fun project to work on, as there was no documentation whatsoever, whether by Google or other existing solutions. So, I documented the key points that explain why a panorama image looks the way it does based on the given inputs (mainly zoom levels).

Other solutions don’t match up because they ignore edge cases, especially pre-2016 images with different resolutions. They used fixed width and height that only worked for post-2016 panoramas, which caused black spaces in older ones.

The way I was able to reverse engineer Google Maps Street View API was by sitting all day for a week, doing nothing but observing the results of the endpoint, testing inputs, assembling panoramas, observing outputs, and repeating. With no documentation, no lead, and no reference, it was all trial and error.

I believe I have covered most edge cases, though I still doubt I may have missed some. Despite testing hundreds of panoramas at different inputs, I’m sure there could be a case I didn’t encounter. So feel free to fork the repo and make a pull request if you come across one, or find a bug/unexpected behavior.

Thanks for checking it out!