Scout system lets low-power devices recognize new wildlife species
Scout: Open-World Species Recognition on the Edge
Computer Vision and Pattern RecognitionDistributed, Parallel, and Cluster ComputingMachine Learning
Summary
Many smart wildlife cameras cannot recognize animals they haven't been trained on and sending all images to the cloud wastes energy and bandwidth. The authors created Scout, which runs mostly on small devices and occasionally asks a powerful cloud AI to identify new species. Scout learns these new species locally, without people having to label them, and keeps using less energy than sending every picture online. This makes it almost as accurate as always using cloud AI but much more efficient for wildlife monitoring.
What this means in practice
- •For wildlife monitoring teams: Deploy camera traps that learn new animal species on-site without manual labels, saving energy and bandwidth while maintaining accuracy.
- •For edge device developers: Build low-power devices capable of updating their visual recognition capabilities by intermittently consulting cloud AI, reducing continuous cloud reliance.
Authors
Mohammad Mehdi Rastikerdar, Hui Guan, Deepak Ganesan
Abstract
Large vision-language models (VLMs) enable recognition beyond a fixed class set, but their computational demands prevent them from running on many edge devices. Cloud offload makes this capability accessible, but sending every image consumes scarce bandwidth and communication energy. We ask how to bring the open-world recognition capability of VLMs to the edge while operating within tight compute, energy, and bandwidth budgets. Wildlife monitoring provides a natural setting for exploring this question because camera traps encounter species not known at deployment. We present Scout, an autonomous open-world recognition system that invokes a cloud VLM intermittently to teach new classes to a compact edge model. Given only the deployment location and empty site frames, Scout autonomously turns each species identified by the VLM into persistent, site-conditioned recognition capability in a resource-efficient edge model, without a predefined species list, human labeling, or manual tuning. Across 30 camera-trap deployments in three regions on an NVIDIA Jetson Orin Nano, the accuracy of Scout remains within 0.1-2.5% of a model given a predefined species list. On species outside its initial class set, Scout achieves 53.7-59.1% accuracy, compared with 56.5-65.1% for full cloud offload, while using 59-71% less deployment energy.