The smarter way to work through camera-trap images.

Yes. The Community plan is free for non-commercial use. It includes AI-assisted labeling and review, occupancy and abundance estimates, export, unlimited image processing, and a Project Model in up to 3 projects. Paid annual plans add Project Models in any number of projects, work across several computers, and commercial use. See Plans.

No. WildObserve™ processes everything on your computer. Your images, their metadata, and your annotations stay there, and we never receive them. The app goes online only to check for updates, activate a license, sign you in with Google if you choose to, and send email alerts if you turn them on. See the Privacy Policy.

Macs with Apple silicon (M1 or later) and macOS 14 or later; Windows 10 or later, from the Microsoft Store; and Linux (Ubuntu 22.04 or later, on a 64-bit Intel or AMD processor). Every plan processes any number of images; how fast depends on your computer.

No. WildObserve™ is a desktop app, so there's no code to write, and it doesn't need a separate graphics card: in tests, it processed 8.1 images a second on an Apple M1 Pro laptop (Tabak 2026, bioRxiv). On Linux, it uses an NVIDIA graphics card if one is set up, and the processor otherwise.

Any species you label. After MegaDetector finds the animals, species suggestions start from SpeciesNet, a general species model from Google. Your Project Model then learns your project's own species, and the names you use, as you label them, including species SpeciesNet doesn't know.

It depends on your images, and the suggestions improve as you review. On a dataset released after SpeciesNet was trained, the Project Model passed SpeciesNet's accuracy with ten labels per species (balanced accuracy 0.783 against 0.751; Tabak 2026, bioRxiv). In the app, Project > Models… shows how well each version of your Project Model did on labels held out from its training.

WildObserve™ runs both for you, with no code, and adds what comes after: a review screen that groups images into capture events, a Project Model that learns your species as you review, occupancy and abundance estimates, and exports in the formats you need. See how it compares with other camera-trap software.

Yes, on every plan. From your reviewed images, it estimates each species' occupancy, allowing for missed detections, and its abundance with the Royle–Nichols model, with site covariates, plots, and tables and a script for R. Its For occupancy list picks which images to review: reviewing about 20% of them this way gives occupancy estimates nearly the same as reviewing all of them (Tabak 2026, EcoEvoRxiv). See the user guide.

CSV (for Excel, R, or Python), Camtrap DP (for GBIF), Wildlife Insights bulk upload, COCO and COCO Camera Traps, YOLO, MegaDetector results, occupancy tables for R, and JSON with full provenance: what the models suggested beside what reviewers saved.

Yes. Everything works offline except checking for updates, activating a license, Sign in with Google, and email alerts. A computer that can't go online can use a license file instead, and if you sign in with Google, you can set a password to sign in without the internet.

Yes. On every plan, each person who reviews on a computer gets their own account, and every record says who confirmed it. With a Lab or Organization plan, projects can also be shared across computers.