No abstract

Facial Expression Recognition (FER) in the wild poses significant challenges due to realistic occlusions, illumination, scale, and head pose variations of the facial images. In this paper, we propose an Edge-AI-driven framework for FER. On the algorithms aspect, we propose two attention modules, Arbitrary-oriented Spatial Pooling (ASP) and Scalable Frequency Pooling (SFP), for effective feature extraction to improve classification accuracy. On the systems aspect, we propose an edge-cloud joint inference architecture for FER to achieve low-latency inference, consisting of a lightweight backbone network running on the edge device, and two optional attention modules partially offloaded to the cloud. Performance evaluation demonstrates that our approach achieves a good balance between classification accuracy and inference latency.

The occurrence status of shale gas inside a nanochannel of a shale reservoir is crucial for shale gas reserve assessment and exploitation. In this work, adopting molecular dynamics simulations, the methane occurrence inside a shale nanochannel was investigated under different formation water saturations. Moreover, employing three common inorganic minerals including of quartz, calcite, and kaolinite, the influence of mineral hydrophilicity on methane occurrence was also examined. The simulations indicate that there are three occurrence statuses of methane inside nanochannels including a free status located in the internal nanochannel, an adsorption status on mineral and water film surfaces, and a dissolution status inside the water phase. Among them, free-status methane is the dominant contribution to the recoverable reserve assessment and could be feasibly exploited. Without formation water, methane gives two occurrence statuses including an adsorption status on mineral surfaces and a free status. Once the formation water emerges, it will preferentially adsorb onto the mineral surface to form a water film. With the increase of formation water saturation, the proportions of free-status and dissolution-status methanes increase, while the proportion of adsorption-status methane decreases. In three mineral nanochannels with hydrophilicity orders of quartz > calcite > kaolinite, the proportion of adsorption-status methane follows the order of kaolinite > calcite > quartz, and free-status methane gives the order of kaolinite ≈ quartz > calcite. The underlying mechanisms of these occurrence features were discussed at length from the view of microscopic interactions among mineral surfaces, water, and methane. Our work presents the methane occurrence structure inside a shale nanochannel, and the discussion of the three methane occurrence statuses could provide fundamental data and theoretical guidance for shale gas reserve assessment and exploitation.

No abstract