Spatio-temporal action detection remains a challenging task due to the inherent coupling between precise spatial localization and effective temporal modeling. Although recent methods have achieved promising performance, they often suffer from spatial feature degradation during temporal aggregation and insufficient modeling of heterogeneous temporal dynamics across actions of varying durations. Moreover, most existing approaches fuse spatio-temporal information in a coarse manner, limiting their robustness in complex scenes. To address these challenges, we propose an efficient spatio-temporal perception network, termed STeP-Net. Specifically, we introduce a frequency-aware feature reconstruction module (FARM) to mitigate the loss of fine-grained spatial details caused by multi-scale temporal processing. Meanwhile, a directional spatial context module (DSCM) is designed to explicitly model local-global spatial dependencies for more robust spatial representation. Furthermore, we propose a heterogeneous temporal reasoning module (HTRM) to capture long-range temporal dependencies while accommodating diverse temporal patterns of different actions. Extensive experiments on three benchmark datasets, including AVA 2.2, UCF101-24, and the self-constructed AVA-E dataset, are conducted to evaluate the proposed method. STeP-Net achieves average mAPs of 18.2% on AVA 2.2 and 72.55% on AVA-E, and reaches 86.97% average mAP on UCF101-24 with a margin of 1.87% over the second-best result, indicating favorable performance across multiple benchmarks.

Splashing is one of the most fascinating liquid phenomena in the real world and it is favored by artists to create stunning visual effects, both statically and dynamically. Unfortunately, the generation of complex and specialized liquid splashes is a challenging task and often requires considerable time and effort. In this paper, we present a novel system that synthesizes realistic liquid splashes from simple user sketch input. Our system adopts a conditional generative adversarial network (cGAN) trained with physics-based simulation data to produce raw liquid splash models from input sketches, and then applies model refinement processes to further improve their small-scale details. The system considers not only the trajectory of every user stroke, but also its speed, which makes the splash model simulation-ready with its underlying 3D flow. Compared with simulation-based modeling techniques through trials and errors, our system offers flexibility, convenience and intuition in liquid splash design and editing. We evaluate the usability and the efficiency of our system in an immersive virtual reality environment. Thanks to this system, an amateur user can now generate a variety of realistic liquid splashes in just a few minutes.