Multi-Scale Motion-Aware Network: enhancing video prediction with lightweight architecture
The proposed MSMANet, a lightweight Multi-Scale Motion-Aware Network that combines implicit spatiotemporal learning with explicit motion decomposition, achieves competitive performance with significantly fewer parameters compared to state-of-the-art models and is suitable for resource-constrained deployment scenarios.