Stage 3. Analyze with AI
What it's for. A camera trapping campaign can generate hundreds of thousands of files. AI filters out empty images, suggests the species and helps recognize individuals. People validate every result before it is used, and that human review is part of the method.
How do we turn thousands of files into information?By the end you will have: Species and jaguars identified and validated by peopleWhat we did in Dzilam: BioScanner
BioScanner is a platform developed by C Minds as part of Tech4Nature Mexico, incubated with the IUCN and Huawei and built on Huawei Cloud. It detects and classifies species in photos and videos, identifies individual jaguars by their spot patterns and generates reports. Its interface is available in Spanish, English, Portuguese and Maya. As of October 2026 it had processed 333,087 photos and videos from the reserve.
It was designed so that non-technical teams can use AI, and with the aim that each community and institution decides about its own data: who sees it, what is shared and with whom.
How BioScanner works
- Organizations and projects
- Communities, governments and organizations create their workspace, invite their team and decide whether each analysis is private or public.
- Cameras on the map
- Cameras are registered on the map, imported from a CSV or read from the files' GPS data. Each folder is assigned to its camera.
- Three AI modules
- Animal detection, species classification and grouping of photos into events and individuals.
- Jaguars
- BioScanner groups records by the animal's flank and suggests matches; people verify them before adding an individual to the catalog.
- Project report
- Species and jaguars by station, independent events, effort, comparison between campaigns, species richness and activity throughout the day.
- On your phone
- A viewing mode brings analyses, galleries and reports to the field. Anyone who knows a species can also contribute photos to improve the models.
Key steps
Create the organization and the project, and invite your team. Start with a small analysis of one or two cameras to get to know the tool.
Register the cameras by importing the CSV from template 3, or let BioScanner read the GPS coordinates from the files. P3
Upload one folder per camera and per visit, set the period and turn on the detection, classification and grouping modules.
Check under “Cameras and dates” that each folder is assigned to its camera and that the camera clocks were set correctly.
Validate the suggested species, taking extra care with rare species, look-alikes and those that matter for your decision. For jaguars, compare both flanks and confirm each new identity with at least two people.
Use the project report and download the validated results to analyze, share or publish them in a standard format such as Camtrap DP.
Lesson learned
AI made it possible to work with an archive of more than 300,000 media files, and expert review gave the results credibility. Students' participation in developing and reviewing models also built new capacity in the region.
Tools to get started
- BioScanner (opens in a new tab) (C Minds). Account-based access, currently in early access.
- Tech4Nature Mexico public repository on GitHub (opens in a new tab), with the notebook for the jaguar identification model.
- Wildlife Insights (opens in a new tab) (Google and partners), a cloud platform that is free for most organizations in the region.
- AddaxAI (opens in a new tab), a free desktop program that runs MegaDetector without coding and works offline.
- Arbimon (opens in a new tab) (getting started (opens in a new tab)) and BirdNET (opens in a new tab), to identify species in acoustic recordings.
Criteria 2.1 and 3.7 (site values and measuring success). Validated records in BioScanner and its reports document key species and make it possible to measure change over time. See how to prepare for the Green List.
Kichwa communities of Napo, Ecuador
Fundación Jocotoco retrained AI models with sounds from the Ecuadorian Amazon and classified more than 2 million audio files collected between 2023 and 2025. Since late 2025 it has operated acoustic monitors in the Kichwa communities of Santa Rita and San Francisco, recognized as Ecuador's first community-based OECMs. The communities use the validated species list to zone their territory, secure funding and present evidence against mining and logging. The results were translated into Kichwa for community assemblies, and the park rangers added the sound of brush cutters to the model.
See report (Mongabay Latam, in Spanish) (opens in a new tab)