 The study presents a new deep convolutional neural network method called Detectory 2 that uses airborne RGB imagery to recognize individual tree crowns in tropical forests. The model was trained on manually delineated tree crowns and evaluated with good accuracy, particularly for tall trees. The approach has many potential applications in forest ecology and conservation, including estimating carbon stocks and monitoring forest phenology and restoration. This article was authored by James G. Siebel, Sebastian H.M. Hickman, Tobias D. Jackson and others.