DAFNe

Overview of DAFNe DAFNe is a deep neural network used for oriented object detection. It is a model that performs predictions on a dense grid over the input image, being architecturally simpler in design as well as easier to optimize compared to its two-stage counterparts. The model reduces prediction complexity by not employing bounding box anchors, which leads to a better separation of bounding boxes especially in the case of dense object distributions. One of the core features of DAFNe is it