Computer vision is a subfield of artificial intelligence (AI) and machine learning (ML) that uses patterns of pixels to extract and classify information from images: in the case of this application, an animal to the species level. A computer vision model is trained through the input of pre-identified (i.e., tagged) images. The number of images required to successfully train a model for the desired accuracy (e.g., F1 Accuracy score of 95% or greater) varies by species but can range from a few hundred to thousands of images (https://wildlifeobserver.net/resources/final-report-phase-ii). Arcadis developed a computer vision model to identify wildlife species from camera trap data utilizing over 19,000 images provided by the Georgia Department of Natural Resources from two separate studies. Image analysis was initially preformed using Microsoft MegaDetector to isolate animal images. The images were then visually reviewed and labeled to the species level and input into the computer vision model to train the software to identify species using pixel pattern recognition. The model exhibits F1 Accuracy of 96% for 10 species when run over 17,000 images. The computer vision model was designed to assess wildlife passage associated with transportation structures, such as bridges and culverts, in Georgia and is focused on species that commonly come into conflict with vehicles when crossing the roadway, such as large mammals. However, the technology has applications for multiple wildlife classes in varying environments. The computer vision model builds on previous applications Arcadis has developed including a ML model using image recognition that can classify species using wildlife passages in real time based on crowdsourced classifications. By allowing the system to automatically detect and classify wildlife, the image recognition reduced cost by diminishing effort for ecologists to determine species resulting in a 90% reduction in workload using large amount of data generated from the online community, and the technology is easily scalable into other datasets.