Automated feature-specific tree species identification from natural images using deep semi-supervised learning

Prior work on plant species classification predominantly focuses building models from isolated attributes. Hence, there is a need for tools that can assist in identification the natural world. We present novel and robust two-fold approach capable of identifying trees real-world setting. Additionally, we leverage unlabelled data through deep semi-supervised learning demonstrate superior performance to supervised learning. Our single-GPU implementation feature recognition uses minimal annotated achieves accuracies 93.96% 93.11% leaves bark, respectively. Further, extract feature-specific datasets 50 by employing this technique. Finally, our method attains 94.04% top-5 accuracy 83.04% bark.