GARNet: Global-Aware Multi-View 3D Reconstruction Network and the Cost-Performance Tradeoff

Deep learning technology has made great progress in multi-view 3D reconstruction tasks. At present, the mainstream solutions adopt different ways to fusion features from several views. Among them, attention-based aggregation function performs relatively well and stably, however, it still an obvious shortcoming strong independence of each view during predicting weights for merging leads a lack adaption global state. In this paper, we propose global-aware approach that builds correlation between branch feature provide comprehensive foundation inference. On basis this, design complete algorithm. Experiments on ShapeNet verify our method outperforms existing SOTA methods. Furthermore, view-reduction based maximizing diversity discuss cost-performance tradeoff model achieve better performance when facing heavy input amount limited computational cost.