Publications

Shape and Texture Aware Graph Processing

Abstract: Neural network training is time-consuming and often application specific. While pre-trained convolutional neural networks can be fine-tuned for other applications, the lower levels are often affected by changes in rotation and scale, since the approach is purely appearance-based. However, hybrid approaches that take both shape and appearance into account (e.g., AAM) have shown success in general computer vision, so in this work we propose a graph network architecture that captures shape and texture information. This separates shape and texture within a graph neural network in a manner analogous to an active appearance model (AAM), offering potential robustness to shape and scale. The training of this graph neural network is done using an unsupervised image data and node position reconstruction approach. This will allow flexible adaptation/fine-tuning to different applications. In addition, we extend the spectral graph convolution approach GCN by automatically generating weights based on a local and global representation in an effort to make the convolutions more flexible and less application or data specific. We call this new approach Generative GCN (GenGCN). Reconstruction experiments in terms of image data and node positions are performed on the CIFAR and MNIST datasets, and a cross database experiment is also conducted.

M. Reale, M. Ochrym, M. Church, N. Goutermout, J. Rubado, M. Cornacchia. “Shape and Texture Aware Graph Processing,” IEEE Applied Imagery Pattern Recognition Workshop (AIPR’20), Oct. 2020.