Facade parsing is a vital technology for applications such as urban modeling, urban planning, and digital twin city construction. High-resolution facade images are particularly important for achieving fine-grained building reconstruction. However, existing facade datasets rarely contain images with resolutions exceeding 2K × 2K. In this paper, we introduce a new dataset featuring much higher-resolution images captured from various angles and with denser window distributions. We also propose a novel method to automatically compute the perspective transformation matrix for generating corrected facade images, which is used to create a twin version of the dataset after perspective correction. Furthermore, we introduce a new network GLNet, designed to achieve superior facade parsing results using high-resolution images as input. Experimental results on three public datasets (CMP, CFP, and ETRIMS) as well as our own dataset demonstrate that GLNet outperform existing methods in facade segmentation. The dataset and code are available at: https://github.com/OctAne0113/GLNet .

Spinal ultrasound imaging offers a radiation-free and cost-effective alternative to X-rays for scoliosis screening, with advantages such as non-ionizing radiation, affordability, and effective soft tissue visualization. These characteristics make it a promising modality for clinical applications, especially in pediatric and frequent monitoring scenarios. However, inherent challenges—such as low contrast, high speckle noise, and various imaging artifacts—significantly degrade image quality, posing difficulties for downstream tasks like anatomical landmark detection. This limitation hinders the deployment of fully automated AI-based diagnostic systems. AI techniques have been applied to spinal ultrasound imaging in previous studies, employing models such as Faster R-CNN, YOLOv8, and YOLOv11 for vertebrae detection. However, challenges in robustness and anatomical prior modeling remain. To address these challenges, we propose a novel framework for the accurate detection of spinal bones in spinal ultrasound images, based on an enhanced YOLOv11 architecture. Our method introduces (1) a heatmap-based center-point branch modeling spinal locations as 2D Gaussians and (2) an Anatomical Spatial Interaction Module that fuses spatial and semantic features via coordinate injection and attention mechanisms. Evaluated on a clinical spinal ultrasound dataset, our approach outperforms these prior methods, achieving a 2.4-point gain in [email protected]:0.95. These results demonstrate the potential of our framework as an effective tool for radiation-free, AI-assisted spinal assessment.

With the proliferation of IoT devices today, the need to strengthen the security of these devices is becoming increasingly urgent, particularly the need to review the security of lightweight block ciphers. SIMON is a lightweight block cipher proposed by the National Security Agency (NSA) of US to provide efficient and secure encryption for resource-constrained devices in IoT systems. This paper aims to evaluate the security of SIMON against mixture differential cryptanalysis, which was proposed in Eurocrypt 2017 to launch the best key-recovery attacks on the most widely used encryption standard AES. Though there have been intensive studies on this cryptanalysis method, its current targets are all aligned block ciphers. Whether the numerous bitwise block ciphers, including SIMON, have weaknesses regarding this method remains unknown. In this paper, we extend the mixture differential cryptanalysis to bit-wise ciphers and develop an SAT-based automatic tool to search for such distinguishers. We interpret the bit-wise mixture differential distinguisher as a variant of differential distinguisher in the multi-key setting with 2-3n as the boundary (n:block size), potentially boosting rounds or improving the signal-to-noise ratio of previous boomerang or classical differential distinguisher. Using SIMON as an example, we discover multi-key distinguishers for up to 17-round SIMON32, 18-round SIMON48, and 23-round SIMON64, which outperform previous results in terms of the number of rounds. This paper reconciles the disparity between mixture differential cryptanalysis applied to word-oriented target ciphers and its application to bit-oriented targets, thereby extending the mixture differential cryptanalysis to a broader range of block ciphers.

Participatory sensing uses both local devices for data collection and cloud-based servers for processing. However, transferring the collected data to the cloud can lead to draining device battery power and cause network bandwidth bottlenecks, especially for large multimedia files. In this paper, we investigate how the processing resources at the edge of the network can be leveraged to enable efficient participatory sensing that avoids heavy network traffic. In particular, we report on the experiences of designing, implementing, and evaluating a sensing system that constructs indoor maps by recognizing door signs. A distinguishing characteristic of our system is an almost exclusive use of edge-based processing for tasks that include ML-based image recognition, human-assisted data verification, data model retraining, and administrative data flow aggregation. Our evaluation shows that our system architecture effectively leverages the available edge resources, while greatly reducing network traffic. Based on our experiences of implementing and evaluating our system prototype, we identify several open research directions for further advancing edge-based participatory sensing.