This demo originates from my group's deep learning project titled "Fine-grained Classification of Bird Species" which won the Intuitive Surgical Best Project Award. We investigated the utility of both orderful and orderless second-order information for fine-grained image classification tasks. To this end we compared the performance of self-attention mechanisms and bilinear convolutional networks. Finally, we proposed a novel architecture which aimed to leverage the dataset's hierarchical labeling to increase predictive context. Here is a brief paper summarizing our approaches and results. The project was implmented with PyTorch.