# Scout: Open-World Species Recognition on the Edge

arXiv: https://arxiv.org/abs/2609.22897 · HTML: https://mlsanity.com/paper/2609.22897/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 52 references that are in the map.
- First posted on arXiv 19 Sep 2026.

## Placed in

- Multimodal LLMs / vision-language models

## Builds on (most cited first, 26 of 26)

- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [Segment Anything](https://mlsanity.com/paper/2304.02643/) (2023)
- [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://mlsanity.com/paper/1905.11946/) (2019)
- [Overcoming catastrophic forgetting in neural networks](https://mlsanity.com/paper/1612.00796/) (2016)
- [A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks](https://mlsanity.com/paper/1610.02136/) (2016)
- [iCaRL: Incremental Classifier and Representation Learning](https://mlsanity.com/paper/1611.07725/) (2016)
- [Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference](https://mlsanity.com/paper/1712.05877/) (2017)
- [Energy-based Out-of-distribution Detection](https://mlsanity.com/paper/2010.03759/) (2020)
- [Pruning Filters for Efficient ConvNets](https://mlsanity.com/paper/1608.08710/) (2016)
- [MnasNet: Platform-Aware Neural Architecture Search for Mobile](https://mlsanity.com/paper/1807.11626/) (2018)
- [Large Language Models as Optimizers](https://mlsanity.com/paper/2309.03409/) (2023)
- [Once-for-All: Train One Network and Specialize it for Efficient Deployment](https://mlsanity.com/paper/1908.09791/) (2019)
- [Recognition in Terra Incognita](https://mlsanity.com/paper/1807.04975/) (2018)
- [Towards Open Set Deep Networks](https://mlsanity.com/paper/1511.06233/) (2015)
- [AnyDoor: Zero-shot Object-level Image Customization](https://mlsanity.com/paper/2307.09481/) (2023)
- [Unified Deep Supervised Domain Adaptation and Generalization](https://mlsanity.com/paper/1709.10190/) (2017)
- [Towards Open World Recognition](https://mlsanity.com/paper/1412.5687/) (2014)
- [BioCLIP: A Vision Foundation Model for the Tree of Life](https://mlsanity.com/paper/2311.18803/) (2023)
- [Adversarial Reciprocal Points Learning for Open Set Recognition](https://mlsanity.com/paper/2103.00953/) (2021)
- [Efficient Pipeline for Camera Trap Image Review](https://mlsanity.com/paper/1907.06772/) (2019)
- [BioCLIP 2: Emergent Properties from Scaling Hierarchical Contrastive Learning](https://mlsanity.com/paper/2505.23883/) (2025)
- [ControlCom: Controllable Image Composition using Diffusion Model](https://mlsanity.com/paper/2308.10040/) (2023)
- [Zero-Shot Deep Domain Adaptation](https://mlsanity.com/paper/1707.01922/) (2017)
- [CATALOG: A Camera Trap Language-guided Contrastive Learning Model](https://mlsanity.com/paper/2412.10624/) (2024)
- [Lessons and Open Questions from a Unified Study of Camera-Trap Species Recognition Over Time](https://mlsanity.com/paper/2603.20509/) (2026)
- [WildFit: Autonomous In-situ Model Adaptation for Resource-Constrained IoT Systems](https://mlsanity.com/paper/2409.07796/) (2024)

## Abstract (arXiv)

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.