Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Sep 5;20(9):e0327969.
doi: 10.1371/journal.pone.0327969. eCollection 2025.

Artificial vision models for the identification of Mediterranean flora: An analysis in four ecosystems

Parminder Kaur  1 , Anna Grassi  2 , Federica Bonini  2 , Barbara Valle  3   4   5 , Marina Serena Borgatti  5 , Giovanni Rivieccio  6 , Agnese Denaro  6   7 , Leopoldo de Simone  4 , Emanuele Fanfarillo  3   4 , Paolo Remagnino  1
Affiliations

Artificial vision models for the identification of Mediterranean flora: An analysis in four ecosystems

Parminder Kaur et al. PLoS One. .

Abstract

Object identification has been widely used in several applications, utilising the annotated data with bounding boxes to specify each object's exact location and category in images and videos. However, relatively little research has been conducted on identifying plant species in their natural environments. Natural habitats play a crucial role in preserving biodiversity, ecological balance, and overall ecosystem health. So, effective monitoring of habitats is necessary for safeguarding them, and one way of doing this is by identifying the typical and early warning plant species. Our study quantitatively evaluates the performance of six popular object detection models on our dataset collected in the wild, comprising various plant species from four habitats: screes, dunes, grasslands, and forests. The dataset employed in this work includes the data collected by human operators and the quadrupedal robot ANYmal C. The pre-trained object detection models have been chosen for experiments, and they are fine-tuned on our dataset to achieve better performance. These models incorporate two one-stage (RetinaNet and YOLOv8n), two two-stage (Faster RCNN and Cascade RCNN), and two transformer-based detectors (DETR and Deformable DETR). Extensive experimentation has been performed on the four habitat datasets by applying class balancing and hyperparameter tuning, and the obtained results are discussed.

PubMed Disclaimer

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Fig 1
Fig 1. The quadruped robot ANYmal C deployed in the screes habitat [19].
ANYmal C is equipped with various sensors, including four Intel RealSense D435 RGB-D cameras for capturing high-resolution images. These sensors enable a large-scale and time-effective data acquisition that can be later used for habitat monitoring.
Fig 2
Fig 2. Sample images from each of the four habitats. Only a few of the species considered are displayed in this figure for brevity.
Fig 3
Fig 3. Total bbox instances (class-wise) in training dataset of all habitats.
Fig 4
Fig 4. Bounding box predictions from six models on a Screes test image.
(A) Ground truth bounding boxes. (B) Faster RCNN. (C) Cascade RCNN. (D) RetinaNet. (E) YOLOv8. (F) DETR. (G) Deformable DETR.
Fig 5
Fig 5. Bounding box predictions from six models on a Dunes test image.
(A) Ground truth bounding boxes. (B) Faster RCNN. (C) Cascade RCNN. (D) RetinaNet. (E) YOLOv8. (F) DETR. (G) Deformable DETR.
Fig 6
Fig 6. Bounding box predictions from six models on a Grasslands test image.
(A) Ground truth bounding boxes. (B) Faster RCNN. (C) Cascade RCNN. (D) RetinaNet. (E) YOLOv8. (F) DETR. (G) Deformable DETR.
Fig 7
Fig 7. Bounding box predictions from six models on a Forests test image.
(A) Ground truth bounding boxes. (B) Faster RCNN. (C) Cascade RCNN. (D) RetinaNet. (E) YOLOv8. (F) DETR. (G) Deformable DETR.
Fig 8
Fig 8. Screes test data confusion matrix of all models.
The numbers in the confusion matrices represent percentages.
Fig 9
Fig 9. Dunes test data confusion matrix of all models.
The numbers in the confusion matrices represent percentages.
Fig 10
Fig 10. Grasslands test data confusion matrix of all models.
The numbers in the confusion matrices represent percentages.
Fig 11
Fig 11. Forests test data confusion matrix of all models.
The numbers in the confusion matrices represent percentages.
See this image and copyright information in PMC

References

    1. Noss RF. Ecosystems as conservation targets. Trends Ecol Evol. 1996;11(8):351. doi: 10.1016/0169-5347(96)20058-8 - DOI - PubMed
    1. Cowling RM, Knight AT, Faith DP, Ferrier S, Lombard AT, Driver A, et al. Nature conservation requires more than a passion for species. Conservation Biology. 2004;18(6):1674–6. doi: 10.1111/j.1523-1739.2004.00296.x - DOI
    1. Bunce RGH, Bogers MMB, Evans D, Halada L, Jongman RHG, Mucher CA, et al. The significance of habitats as indicators of biodiversity and their links to species. Ecological Indicators. 2013;33:19–25. doi: 10.1016/j.ecolind.2012.07.014 - DOI
    1. Berg C, Abdank A, Isermann M, Jansen F, Timmermann T, Dengler J. Red lists and conservation prioritization of plant communities – a methodological framework. Applied Vegetation Science. 2014;17(3):504–15. doi: 10.1111/avsc.12093 - DOI
    1. Keith DA, Rodríguez JP, Brooks TM, Burgman MA, Barrow EG, Bland L, et al. The IUCN red list of ecosystems: motivations, challenges, and applications. Conservation Letters. 2015;8(3):214–26. doi: 10.1111/conl.12167 - DOI

LinkOut - more resources