What this costs, and how it is actually going
Generated from the live database on
2026-10-06 11:45. Every number
here is produced by tools/support_page.py reading the database, not
typed in by hand.
Chip in — $sparrowmap on Cash App
Scan the code, or tap the address once to select all of it.
47uEmSACPB9YpdEUG99jnwAUKVBAEQtKH7EvU95Yxa1HcC7g5CefBDxexWKrLEjDEodZwT24HVQ61NoYeqaadQHMKYJ6JN6
The volunteer network
The number we are not going to inflate. The database also holds 19,890 public traffic cameras, 15,116 of them reporting. Those are not volunteers - this project found them and enrolled them itself, and they are counted separately everywhere on this page. Adding them to the number above would make the network look thirty times larger than it is.
And the number that actually matters. Of the 2,839 volunteer cameras enrolled more than three days ago, 12 were still reporting in the last 24 hours. That is 0%. Reach has not been the problem; people set a camera up and stop. That is the real constraint on this project and it would be dishonest to leave it off a page asking for money.
Cameras enrolled per day, people only
Output
Most of what the cameras see is ordinary traffic, and that is deliberate: those rows carry no plate text, are not searchable, and exist only so the map can show that a road is busy. The government-vehicle count is the small one because publishing one takes a human confirming it.
What this needs, and what each piece unlocks
Rather than asking for support in the abstract, here is the actual hardware this needs and what each piece would answer. Prices checked 2026-08-19, not remembered: they moved a long way this year.
The single highest-leverage item on this list. Nobody has confirmed whether the government-vehicle classifier converts to this accelerator at a useful speed, and everything about a self-contained node depends on the answer. If it works, a camera can decide for itself instead of sending a crop home. If it does not, that is worth knowing for $70 rather than guessing about for another year.
The reference build to test the above on, and the board volunteers are most likely to own. Pi 4 is marginal and a Zero cannot do it. Testing on hardware nobody has means the instructions are guesses.
Runs the detector and the video work locally instead of renting GPUs by the hour. Two things follow: the running cost of every experiment drops to electricity, and product photos never leave the building, because generation happens here rather than on somebody else's cloud.
$0 of $1,338 toward the list. Nothing received toward it yet. There is no automatic read of the donation balance, so this is updated manually rather than shown as something it is not.
After that, and deliberately unpriced
- Solar kit for an outdoor node A 100W panel and a 100Ah battery to survive a Michigan January on a pole. Deliberately unpriced here: the last figure I had was an estimate and I have not re-checked it, so it is not going on the page as a number.
If none of it arrives, the project carries on. Everything above makes it faster or answers a question sooner; none of it is keeping the lights on. The running costs below are the part that does that.
What it is running on right now
What it costs to run
| What | Per month |
|---|---|
| Map server and traffic-camera reader Helsinki. 8 cores, 32 GB, 240 GB disk. One machine doing both jobs since 2026-08-18: it serves the map, takes every camera upload, holds the database, and reads several thousand public traffic cameras on a cycle. It is nearly the whole bill. | $162.99 |
| Traffic-camera reader, US Ashburn, Virginia. A second, smaller reader placed in the United States so that American traffic cameras are read from inside the country rather than across the Atlantic. | $37.49 |
| Domain sparrowmap.com. Paid monthly by the operator, not through the hosting console, which is why it is not on the Hetzner invoice with the servers. | $165.84 |
| Total | $366.32 |
Last updated 2026-08-19. Taken from the per-month price each server shows in the hosting console, not estimated. The first month has not closed yet, so these are not invoiced figures.
367 people at $1 a month covers all of it. That is the whole funding model. Not a subscription, not a tier, no feature held back for payers - the map is public and stays public either way.
Chip in — $sparrowmap on Cash App
Scan the code, or tap the address once to select all of it.
47uEmSACPB9YpdEUG99jnwAUKVBAEQtKH7EvU95Yxa1HcC7g5CefBDxexWKrLEjDEodZwT24HVQ61NoYeqaadQHMKYJ6JN6
If money ever exceeds what the servers cost, the surplus goes to the same place: more polling capacity and bandwidth. There is no salary in this.
What your money does not buy
No access to anything a visitor cannot see. No private tier, no early sightings, no plate lookups, no removals. The two-tier design is not a paid feature and cannot be bought around: a private vehicle's plate is painted out on the camera before it is uploaded, and there is nothing on the server to sell.
Transparency · Checksums · Status · Source