Scarce data driven deep learning of drones via generalized data distribution space

Increased drone proliferation in civilian and professional settings has created new threat vectors for airports national infrastructures. The economic damage a single major airport from incursions is estimated to be millions per day. Due the lack of balanced representation data, training accurate deep learning detection algorithms under scarce data an open challenge. Existing methods largely rely on collecting diverse comprehensive experimental footage artificially induced augmentation, transfer meta-learning, as well physics-informed learning. However, these cannot guarantee capturing designs fully understanding feature space drones. Here, we show how general distribution via generative adversarial network (GAN), explaining under-learned features using topological analysis (TDA) can allow us acquire under-represented achieve rapid more We demonstrate our results image dataset, which contains both real images simulated computer-aided design. When compared random, tag-informed expert-informed collections (discriminator accuracy 94.67%, 94.53% 91.07%, respectively, after 200 epochs), proposed GAN-TDA-informed collection method offers significant 4% improvement (99.42% epochs). believe that this approach exploiting knowledge neural networks applied wide range challenges.