Flock Cameras & Surveillance in Acadiana

The map

Each wedge is a plate reader, pointing the way it faces — the direction of traffic it reads. Circles are gunshot detectors and any plate readers with no recorded bearing. Tap any device for details and a link to its OpenStreetMap record.

Where they are

Who makes them

Vendor comes from the contributor-supplied brand, manufacturer, and operator tags. Where all three are blank the camera is counted but not attributed.

Who runs them

The map shows you where the cameras are. It can’t tell you whose they are — 193 of the 208 plate readers here carry no operator tag at all. Two agencies have said on the record that some of them are theirs.

That accounts for a fraction of the map. Who operates the rest isn’t in any public record we could find. The survey behind this page records an operator for only 15 of the 208 readers — and 12 of those just repeat the reader’s own vendor name rather than naming an agency. Only 3 name an actual agency.

What they cost

Almost nothing about the price of this is public. That is itself the finding.

• No contract, no figure. LPD confirmed it owns the readers but did not provide The Advocate details of a contract or the associated costs.
• Even the line-item detail doesn’t break them out. The Recap in LPD’s adopted FY 2025-26budget only goes down to categories — purchased services, materials and supplies, capital outlay. The Detail goes further, but only into operating accounts like SUP & MAT-IN CAR CAMERA and EQUIP MAINT-NEIGHBORHD CAMERAS; its capital outlay is a single lump-sum line, not itemized. The individual purchases — CRIME SCENE CAMERAS NEW, PARK CAMERAS NEW-40 — show up only in a separate capital appropriations summary later in the book. None of it says plate reader: across all 23,370 lines of the adopted budget, the words flock, plate reader, license plate, ALPR, gunshot, raven, and shotspotter never appear.
• The nearest number covers all cameras. LPD reports, in its own summary of the prior year in that book, that it obtained $550,000 in camera funding and installed 315 additional camera views. It also tracks a performance measure it labels “Cameras.” Neither separates plate readers from ordinary cameras.
• One contract is documented.Gunshot detection: $25,000 for a year of coverage over one square mile, October 2024 to October 2025. Lt. Brad Robin told The Advocate the money came out of the department’s surveillance camera budget and was signed off on by LPD command staff, then-CAO Cydra Wingerter and then-Mayor-President Josh Guillory — an administrative sign-off.

Whether that gunshot contract was renewed past October 2025 isn’t something we could establish from public documents.

What happens to the data

A plate reader doesn’t watch. It records — the plate, the time, the place, and a description of the vehicle — and files that in a searchable database.

• Thirty days, then deleted. LPD told KLFY images are automatically deleted after 30 days. ULPD holds its campus data for the same 30 days.
• Searches are logged.LPD says only authorized officers can access the system and that every search is logged and subject to review. That is the department describing its own controls. We found no outside audit of Lafayette’s search logs.
• Vehicles, not people, is the department’s answer.Asked about privacy, Richard told KLFY: “It’s not tracking people… it doesn’t identify the people that are in the vehicles. All it’s doing is helping us identify the vehicle.”
• The network reaches past the cameras.The Advocate reported that Virginia media outlets found a local police department there sharing data with federal immigration authorities despite that agency’s claim otherwise, and that in August 2025 Flock said it did not have a contractual relationship with U.S. Department of Homeland Security agencies, but that it had engaged in “limited pilots” with Customs and Border Protection and Homeland Security Investigations.
• The courts have not settled it. The ACLU argues that letting police run warrantless searches of Flock databases violates the Fourth Amendment. A federal district court in Virginia went the other way in October 2024, upholding a warrantless Flock search used to find a robbery suspect.
• The Legislature may take it up.Rep. Danny McCormick of Caddo Parish says he is drafting a bill to ban Flock cameras that capture and log plate data in searchable databases. “Warrantless mass surveillance is what these flock cameras do,” he told Louisiana Radio Network in August 2026. “They record the license plates. They can share them without warrants.”

Gunshot detection

Lafayette’s gunshot detection isn’t ShotSpotter. It’s Raven, built by Flock — the same company as the plate readers.

LPD signed up to test Raven in December 2021 and had 60 solar-powered microphones installed between August and October 2023, covering one square mile across two zones in Precinct 1. This map shows 22 gunshot detectors in total, four of them in St. Martin Parish — outside the city LPD polices, where nothing we found says who runs those — leaving 18, not 22, as the figure to weigh against LPD’s 60. Either way, it is a fair measure of how far a crowdsourced survey sits from a complete one.

Lt. Brad Robin told The Advocate on January 31, 2024 that officers had received 55 alerts in the previous 30 days. To his knowledge there had been one confirmed false positive — metal repeatedly striking metal — and one shooting the sensors missed, which happened inside a vehicle beside a sensor. “Instead of driving around trying to figure out where these shots are, it tells you a more precise location,” he said. Treat that as an anecdote rather than a record: Robin also said the department was not keeping firm data about the system and its alerts, including whether an alert turned out to be gunfire at all.

Flock told GovTech in 2021 that it put Raven’s accuracy “north of 90 percent” and planned to rely solely on its AI algorithms, where ShotSpotter routes alerts through human reviewers. Whether that is still how Raven works, we can’t say from the record.

Both of those hold at once. Faster and better-aimed dispatch is a real thing the technology does, and the Kansas City numbers show it. Fewer shootings is a different claim, and the research hasn’t been able to show it.

One caveat sits under all of it. That research is about ShotSpotter and systems like it, not about Raven. When The Acadiana Advocate looked in 2024 it could not find long-term data or outside studies on Raven’s efficacy, which it put down to how recently the product reached the market. Nothing we found since has changed that. So the honest read is that the technology Lafayette bought has not been independently evaluated, and the closest evidence available is about a competitor.

How this data is made

Every device on this map comes from OpenStreetMap, the open map database anyone can edit. Volunteers — many organised through DeFlock, which maps plate readers nationally — spot cameras in the field and record them. We query that database directly for the nine parishes and publish the result as a dated snapshot.

• It is an undercount. Nobody has systematically surveyed every road in Acadiana. Absence of a marker is not evidence of absence of a camera.
• It is a deliberate undercount, too.This map plots only automatic plate readers and gunshot detectors. OpenStreetMap also records other surveillance cameras across Acadiana — doorbell cameras, business CCTV, and the like — and we filter every one of them out by design. On a page arguing that agencies won’t say what they’re running, our own exclusion belongs on the record too.
• Positions are contributor-reported. We have not field-verified them.
• Vendor attribution is patchy. Contributors record it inconsistently, and some cameras carry no vendor at all.
• It can blend more than one server’s data. Our query retries against three separate Overpass mirrors when one is slow or unavailable, and mirrors replicate OpenStreetMap at different lags — so a single published snapshot can combine results pulled from up to three different moments.
• It is a snapshot, not a live feed. This data is current as of August 11, 2026.