# Deep learning-based ecological analysis of camera trap images is impacted by training data quality and quantity

arXiv: https://arxiv.org/abs/2408.14348 · HTML: https://mlsanity.com/paper/2408.14348/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 77 references that are in the map.
- First posted on arXiv 26 Aug 2024.

## Builds on (most cited first, 17 of 17)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [A ConvNet for the 2020s](https://mlsanity.com/paper/2201.03545/) (2022)
- [How transferable are features in deep neural networks?](https://mlsanity.com/paper/1411.1792/) (2014)
- [Recognition in Terra Incognita](https://mlsanity.com/paper/1807.04975/) (2018)
- [Deep Learning is Robust to Massive Label Noise](https://mlsanity.com/paper/1705.10694/) (2017)
- [Efficient Pipeline for Camera Trap Image Review](https://mlsanity.com/paper/1907.06772/) (2019)
- [The iWildCam 2021 Competition Dataset](https://mlsanity.com/paper/2105.03494/) (2021)
- [Context R-CNN: Long Term Temporal Context for Per-Camera Object Detection](https://mlsanity.com/paper/1912.03538/) (2019)
- [A deep active learning system for species identification and counting in camera trap images](https://mlsanity.com/paper/1910.09716/) (2019)
- [SVL-Adapter: Self-Supervised Adapter for Vision-Language Pretrained Models](https://mlsanity.com/paper/2210.03794/) (2022)
- [Iterative Human and Automated Identification of Wildlife Images](https://mlsanity.com/paper/2105.02320/) (2021)
- [ConvNet vs Transformer, Supervised vs CLIP: Beyond ImageNet Accuracy](https://mlsanity.com/paper/2311.09215/) (2023)
- [Focus on the Positives: Self-Supervised Learning for Biodiversity Monitoring](https://mlsanity.com/paper/2108.06435/) (2021)
- [The Resistance to Label Noise in K-NN and DNN Depends on its Concentration](https://mlsanity.com/paper/1803.11410/) (2018)
- [Comparison between transformers and convolutional models for fine-grained classification of insects](https://mlsanity.com/paper/2307.11112/) (2023)

## Built on by (most cited first, 3 of 3)

- [Lessons and Open Questions from a Unified Study of Camera-Trap Species Recognition Over Time](https://mlsanity.com/paper/2603.20509/) (2026)
- [Camera trap classification with deep learning under ground truth uncertainty](https://mlsanity.com/paper/2608.30789/) (2026)
- [Targeted Review for AI-Assisted Biodiversity Surveys: Active Continuous-Score Occupancy Modeling](https://mlsanity.com/paper/2609.25657/) (2026)

## Abstract (arXiv)

Large image collections generated from camera traps offer valuable insights into species richness, occupancy, and activity patterns, significantly aiding biodiversity monitoring. However, the manual processing of these datasets is time-consuming, hindering analytical processes. To address this, deep neural networks have been widely adopted to automate image labelling, but the impact of classification error on key ecological metrics remains unclear. Here, we analyse data from camera trap collections in an African savannah (82,300 labelled images, 47 species) and an Asian sub-tropical dry forest (40,308 labelled images, 29 species) to compare ecological metrics derived from expert-generated species identifications with those generated by deep learning classification models. We specifically assess the impact of deep learning model architecture, proportion of label noise in the training data, and the size of the training dataset on three key ecological metrics: species richness, occupancy, and activity patterns. We found that predictions of species richness derived from deep neural networks closely match those calculated from expert labels and remained resilient to up to 10% noise in the training dataset (mis-labelled images) and a 50% reduction in the training dataset size. We found that our choice of deep learning model architecture (ResNet vs ConvNext-T) or depth (ResNet18, 50, 101) did not impact predicted ecological metrics. In contrast, species-specific metrics were more sensitive; less common and visually similar species were disproportionately affected by a reduction in deep neural network accuracy, with consequences for occupancy and diel activity pattern estimates. To ensure the reliability of their findings, practitioners should prioritize creating large, clean training sets and account for class imbalance across species over exploring numerous deep learning model architectures.