Wildlife preservation requires careful monitoring of populations which traditionally requires performing time-intensive censuses in the field. When these populations reach in the thousands of individuals, this work can be particularly complex unless researchers start relying on statistical estimates. Object counting in images can provides a means to monitor populations easily.\ This project first attempts counting wildlife, seals in this instance, using a simple K-Means clustering using the raw data, tuning for different hyperparameters (namely type of distance) then adding bounding boxes to the training data set for compatibility with Facebook's Detectron2 object detection platform.
To see code and results, follow the link to my Github page.