Hand-annotated ~3000 images of various robotics field elements, including cones and cubes, using Make Sense AI. Fed image data into computer vision YOLOv5 model for training through TensorFlow framework. Model achieved 95% accuracy score. Exported model to .tflite format for use in real-time camera detection on the field.