TDWG2026
Home / Talks / SAL A
talk · Tuesday 22 September · 09:57–⁠10:09 · SAL A

Trapper 2.0: scalable, open-source platform for managing camera trapping projects with integrated AI pipelines

Karolina Kuczkowska · AI for Biodiversity Data

Recording time 1:22:54–1:41:14Open on Vimeo ↗

Summaries and key moments are machine-produced. Check the recording for the speaker’s exact words.

The short versionTrapper 2.0 offers an open, model-agnostic, edge-deployable pipeline for camera trap photos and videos with expert review and Camtrap DP export.

Overview

What this was about

Karolina Kuczkowska presented Trapper, a ten-year-old open-source data infrastructure for camera trapping projects, and its forthcoming version 2, which adds video analysis. Its customisable AI pipeline chains object detection (MegaDetector or YOLO models), species classification, multi-object tracking with custom post-processing, pose estimation and, in development, depth estimation for population density methods, with experts correcting results in the interface. New models can be added via a form or a Python connector class. Trapper runs wherever Docker runs, from cloud servers to NVIDIA Jetson or Raspberry Pi; the WildLabs-funded Trapper Keeper field box enables edge processing, which she argued is key for hotspots without infrastructure, data security and lower environmental impact. Data export in Camtrap DP allows publication to GBIF via the IPT, with direct publication planned.

Why it matters. Manual camera trap analysis can delay results by a year or more; an open, scalable pipeline that runs even offline in the field brings automation to where biodiversity is.

Key ideas

In the room

  • Trapper is a suite of tools (GUI, back end, AI components, command-line tools), fully open source on GitLab.
  • Version 2 supports videos and is planned for official release before the end of the year.
  • Video counts need tracking to determine whether detections in different frames are the same animal.
  • Pose tracking currently improves tracking accuracy; behaviour models are a planned next step.
  • Depth estimation, calibrated with objects at known distances, produces per-frame distance maps for population size estimation.
  • The full stack including AI runs on CPU, GPU or Hailo accelerators on microservers.
  • Over 40 million images processed by partner institutions; Slack community of almost 200 experts.
Jump in

Notable moments

In their words

Transcript

Automatically generated captions can contain mistakes, especially in names and technical terms. Times are relative to the room recording.

Read the transcript ↓
Loading transcript…
Keep exploring

Related talks.

See the map ↗
talk · 24 Sept · 14:15–⁠14:30 · SAL C

Camtrap DP: extending data standard designed for camera trapping research to support AI workflows and new data sources

Camtrap DP fits data from any stationary sensor, and converging community feedback shows how to extend it for AI provenance and multimodal monitoring.

Karolina Kuczkowska
talk · 22 Sept · 11:15–⁠11:27 · SAL A

The WildLIVE Portal: Towards Scalable Wildlife Monitoring through the Integration of Citizen Science and Human-in-the-Loop AI Workflows

WildLIVE combines a model registry, provenance-tracked human verification and FAIR publication to move camera trap data from raw images to publishable data.

Rajapreethi Rajendran
talk · 22 Sept · 09:09–⁠09:21 · SAL A

Scalable Edge AI for Real-Time Biodiversity Monitoring: tracking invasive plant species in roadside imagery

Distilling a tiling-based plant identifier into a single ConvNeXt model makes high-resolution roadside invasive-species monitoring fast, cheap and more accurate.

Giulio Martellucci
Programme abstract

Trapper 2.0: scalable, open-source platform for managing camera trapping projects with integrated AI pipelines

Karolina Kuczkowska

Camera trapping has become a powerful tool for wildlife monitoring, but the resulting multimedia datasets, often comprising millions of images and videos, present significant analytical challenges. Manual annotation is time-consuming, expensive, and difficult to scale, making AI integration essential for modern biodiversity monitoring workflows.

Trapper is an open-source, web-based platform designed to manage camera trap data at any scale, from individual research projects to large citizen science initiatives. It allows users to build automated AI analysis pipelines consisting of a variety of open-source AI models used for object detection, tracking objects in videos, species classification and distance and movement speed detection. Trapper’s architecture allows for easy integration of new models as they are being developed.

The results of AI analysis can be reviewed and corrected by human experts. This human-in-the-loop approach maximises annotation efficiency without sacrificing data quality. The recently released version, Trapper 2.0, features support for processing videos as well as images. Trapper is compatible with the Camtrap DP data standard, ensuring interoperability with GBIF and broader biodiversity data infrastructures.

An important feature of Trapper’s design is deployment flexibility. The entire Trapper stack, including the AI processing components, can be run on a personal computer, a local server or a cloud server, as well as on microservers like Raspberry Pi. AI processing can be done using CPU alone, with GPU acceleration or using Hailo-based processing, enabling computation on edge devices. This helps to significantly decrease computation costs, making AI-assisted biodiversity monitoring more accessible even with limited resources.

Submitted to the conference programme before the talk, so the talk itself may differ.