Seg-HGNN: Unsupervised and Light-Weight Image Segmentation with Hyperbolic Graph Neural Networks
This work addresses image segmentation for computer vision applications, offering an incremental improvement in unsupervised methods with a focus on efficiency and small model size.
The paper tackled image segmentation by proposing a light-weight hyperbolic graph neural network, achieving improvements of 2.5% and 4% on VOC datasets for localization and 0.8% and 1.3% on CUB-200 and ECSSD for segmentation over the best unsupervised method, with under 7.5k parameters and processing about 2 images per second on standard GPUs.
Image analysis in the euclidean space through linear hyperspaces is well studied. However, in the quest for more effective image representations, we turn to hyperbolic manifolds. They provide a compelling alternative to capture complex hierarchical relationships in images with remarkably small dimensionality. To demonstrate hyperbolic embeddings' competence, we introduce a light-weight hyperbolic graph neural network for image segmentation, encompassing patch-level features in a very small embedding size. Our solution, Seg-HGNN, surpasses the current best unsupervised method by 2.5\%, 4\% on VOC-07, VOC-12 for localization, and by 0.8\%, 1.3\% on CUB-200, ECSSD for segmentation, respectively. With less than 7.5k trainable parameters, Seg-HGNN delivers effective and fast ($\approx 2$ images/second) results on very standard GPUs like the GTX1650. This empirical evaluation presents compelling evidence of the efficacy and potential of hyperbolic representations for vision tasks.