The rapid advancement of DeepFake technology poses a formidable challenge to information security. Existing multimodal detection methods have achieved success by examining audio-visual consistency, but they primarily model features in the spatial or high-level semantic domains, overlooking the rich information in the temporal-frequency domain. To address this, we propose a novel, plug-and-play framework named FEMC (Frequency-Enhanced Multi-Modal Consistency Learning). FEMC operates in parallel with arbitrary base models and introduces frequency-domain analysis branches to capture signal-level inconsistencies. The framework jointly models two levels of consistency. First, it learns intra-modal temporal coherence by applying an enhanced attention mechanism to frequency-domain representations. Second, it enforces inter-modal temporal correspondence through a cross-modal attention module and a dedicated contrastive loss. Extensive experiments on the dataset FakeAVCeleb validate our approach. Integrating the FEMC module yields significant performance gains across diverse base models, boosting AV-HuBERT’s accuracy by 5.39% and a combined R(2+1)D+VGGish model’s by 3.08%, while establishing a new state-of-the-art AUC of 99.76%.

In recent years, many studies proposed to generate solutions for Directed Acyclic Graph (DAG) scheduling problem in one shot by combining reinforcement learning and list scheduling heuristic. However, these existing methods suffer from biased estimation of sampling probabilities and inefficient guidance in training, due to redundant comparisons among node priorities and the sparse reward challenge. To address these issues, we analyze of the limitations of these existing methods, and propose a novel one-shot DAG scheduling method with comparability identification and dense reward signal, based on the policy gradient framework. In our method, a comparable antichain identification mechanism is proposed to eliminate the problem of redundant nodewise priority comparison. We also propose a dense reward signal for node level decision-making optimization in training, effectively addressing the sparse reward challenge. The experimental results show that the proposed method can yield superior results of scheduling objectives compared to other learning-based DAG scheduling methods.

Effective multiscale medical image segmentation requires simultaneously preserving smooth spatial continuity and accurately delineating high-frequency boundaries, yet pixel-wise decoders often fail to maintain this balance consistently across varying resolutions. We introduce GauSAM, which seamlessly integrates contour‑guided 2D Gaussian probability fields into the Segment Anything Model to address these challenges. In our framework, segmentation masks are parameterized as continuous probability fields of learnable 2D Gaussian primitives, enforcing spatially smooth and structurally consistent. Contourlet transforms extract rich multidirectional frequency information, notably edges and fine textures, which dynamically guide the spatial distribution of Gaussian primitives to substantially improve boundary fidelity in complex structures. The incorporation of these high-frequency contour priors also enriches the expressive capacity of the SAM image encoder. Extensive experiments on diverse 2D medical segmentation tasks confirm that GauSAM consistently delivers robust generalization and state-of-the-art performance with only 1.2M trainable parameters. The official implementation of GauSAM is publicly available at https://github.com/Quinten-Wu504/GauSAM.

In the era of smart grid development, it is an important task to improve the efficiency and intelligence of the grid control system. One of the problems that need to be solved is the tedious and error-prone manual drawing of the CAD design diagram, which requires maintainers to refer to the original electrical diagram for manual drawing. This is a huge workload and is prone to mistakes. To solve this problem, the automatic identification of electrical component in diagrams is particularly critical, and existing methods are inadequate in terms of algorithmic accuracy and robustness. To achieve better automatic detection, we propose an improved Faster-RCNN algorithm to detect electrical components in diagrams. In order to further improve the recognition efficiency for a multiple scales of electrical components, we replaced the original feature extractor VGG-16(Visual Geometry Group 16-layer model) with ResNet-50 as the feature extractor and introduce a feature fusion network based on attention block to strengthen the detection ability of multi-size electrical components. In the RPN network, the k-means++ algorithm is introduced to better generate anchor boxes. To overcome the problem of information loss caused by directly reducing the original grid drawing to the predetermined size, the overlapping sliding window is used to detect the high-resolution electrical diagram.