This project developed an artificial intelligence system based on a convoluted neural network capable of detecting and counting sea duck individuals in aerial photos and classifying individuals to species and sex when possible, to reduce time and cost commitments associated with processing imagery from aerial surveys. We used a dataset consisting of 810 aerial images containing sea ducks and other birds in offshore and coastal environments. Images were collected from fixed wing aircraft at varying flight heights and image resolution. A biologist reviewed each image identifying and annotating the position of objects of interest (birds) in the image by drawing a bounding box surrounding each object, and assigned species and sex when possible. This dataset was used to train and evaluate model versions and calculate final model performance. We developed an R package (with technical documentation) containing functions that apply the computer vision system, as well as an HTML-based user interface to those functions.