Detection
Find animals in each image or video frame — skip empty shots and focus on captures that matter.
Wildlife AI
From camera trap to species — automatically
The Untamed Data Platform runs its own wildlife AI. It detects animals and identifies the species in every capture from your cameras. Uncertain results go to a person for review. All of it runs on European infrastructure, next to your data.
Capabilities
What the pipeline does
Detection, classification, and human review in one continuous flow.
Species classification
Identify the species behind each detection, with confidence scores you can filter and export.
Human review
Uncertain identifications land in a review queue — confirm or correct before they enter your reports.
In depth
Inside a single capture
Two stages, not one guess. Here is what happens to a single photo between the camera and your dashboard.
-
01
Find the animal
A detector draws a box around each animal in the frame and skips the empty triggers — wind, rain, a passing shadow.
-
02
Name what it is
A classifier reads each detection and predicts the species, with a confidence score attached.
-
03
Coarse when unsure
Sure it is a deer but not which deer? It labels the group — “deer” — instead of guessing a species. A right group beats a wrong species.
-
04
Uncertain goes to a human
Anything below the confidence threshold lands in the review queue, where a person confirms or corrects it.
No black box. Every detection is stamped with the model name and version that produced it — so results stay auditable and reproducible, and the model can be upgraded without touching the data you have already collected.
On the roadmap: identifying each animal separately in busy frames, video tracking that counts a single animal once, and our own European-trained model from the funded AI Wildlife Box.
Pipeline
How wildlife AI fits your workflow
No manual tagging step. Footage from connected cameras or manual upload flows straight into detection, classification, and your dashboard.
-
01
Upload
Photos and videos arrive via FTPS (preferred) or plain FTP from your cameras, or through manual upload from the field.
-
02
Detect & classify
Our inference service finds animals and assigns a species label with a confidence score for each capture.
-
03
Review if needed
Low-confidence results go to the review queue. High-confidence detections move straight to your dashboard.
Infrastructure
Wildlife AI on European infrastructure
The Untamed Data Platform runs its own dedicated inference service for camera-trap workflows — on the same European infrastructure as your footage. Every detection shows species, confidence, and source media. Human in the loop: AI assists identification; uncertain results go to a review queue, and corrections you make are stored for audit and quality tracking.
For the full trust story — hosting, ownership, and platform independence — see Why Untamed Data.
Media
Photos and videos
The pipeline handles standard camera-trap photos (JPEG, PNG) and videos (MP4, MOV, AVI). For video clips, key frames are sampled across the full duration — so you get representative detections without processing hundreds of near-identical frames from the same animal passing through.
Most wildlife cameras are motion-only, so they fire on wind, branches, and warm air just as readily as on an animal. The pipeline drops those empty and false triggers early — so your dashboard highlights real activity: foxes, deer, boar, badgers, and the rest of your local fauna, not another thousand frames of moving grass.
Quality control
Review queue
Not every capture is equally clear. Night shots, partial animals, or unusual angles can lower the model's confidence. When a detection falls below the confidence threshold, it is held in the review queue until a team member confirms or corrects the species.
- Confirm as-is — accept the AI suggestion when it looks right.
- Correct the species — pick the right label from the list or type a new one.
- Bulk actions — confirm or reject multiple detections at once during busy field seasons.
Reviewed detections flow into charts, maps, gallery views, and CSV, PDF, and Camtrap DP exports — the same data path as high-confidence automatic identifications. Every correction is stored with the detection for audit and quality tracking.
Region
Built for European wildlife monitoring
The classification model is tuned for broad wildlife recognition with regional context — well suited to European camera-trap projects from rewilding reserves and forestry estates to research plots and farmland corridors.
Species appear in your dashboard with readable names and filterable counts per project, camera, and date range — ready for reporting to stakeholders, funders, or publication.
Compare
AI vs manual review
How automated species identification compares to fully manual tagging.
| Manual tagging | Untamed Data wildlife AI | |
|---|---|---|
| Speed | Hours per SD card | Automatic on upload |
| Consistency | Varies by reviewer | Same model, same scoring |
| Quality control | 100% human effort | Review queue for uncertain shots |
| Scale | Hard to scale across sites | Multi-camera, multi-project dashboards |
| Data location | Depends on your tools | European-hosted, export anytime |
Next step
See wildlife AI in action
Test the platform, connect a camera, and let species identification run automatically. Or talk to us about larger deployments.
