Automated camera trap species recognition made easy: Using entry-level hardware and few training data

Computer vision methods used to analyse camera trap photos are usually computationally expensive, require large training datasets and typically focus on only one species per photograph or rely on static backgrounds between sequential images. In contrast, our proposed method requires only an entry-level computer and relatively few training data while handling multi-species photos with changing backgrounds. It is able to distinguish between four large mammal species common to the Iona–Skeleton Coast TFCA, namely giraffe, impala, oryx and zebra. Trained on readily available online images and applied to 4000 camera trap photos, the system yielded a recall of 59.1% in detecting the presence of animals in camera trap photos. Precision in detecting animals was 100% while precision in distinguishing between the four species of interest, namely giraffe, impala, oryx and zebra, was 96.8%. Based on the results, the method could be used to filter large raw datasets for photos containing animals, and to label or pre-label photos by species for further analysis. This may make it useful to aid in compiling species inventories, document animal migration, map species distributions and estimate densities.

Publish DateSeptember 2, 2021
Last UpdatedSeptember 2, 2021