Despite notable progress in cross-view localization, extreme viewpoint differences between aerial and satellite imagery remain challenging. Existing methods often struggle to balance retrieval efficiency, matching accuracy, and model lightweightness. We propose a lightweight two-stage framework integrating efficient global retrieval with semantically-enhanced local matching. First, we adapt ternary quantized transformers to cross-view scenarios via progressive transfer learning, boosting retrieval without large-scale pretraining. Second, we introduce a dual-branch architecture that fuses texture and semantic features through cross-attention reasoning, employing semantic conditioning to filter mismatches from repetitive structures. Results on public datasets demonstrate improved localization accuracy and reduced computational cost, validating our framework's suitability for resource-constrained UAVs.

Nowadays, autonomous vehicles (AVs) have become an excellent solution for alleviating the burden of the traffic system and improving safety. However, the large-scale application of AVs is significantly restrained by the high-power cost of their numerous sensors and processors, which can reach hundreds of watts and occupy most of the vehicular power consumption. To meet such tremendous computation requirements, most AVs rely on vehicle-to-everything (V2X) communication with mobile edge computing (MEC) techniques to relocate their tasks to external computing. In this paper, the problem of optimizing the task and path assignment process in the V2X-MEC system is tackled for further reducing vehicular power consumption with the constraint of transmission latency. We incorporate the Nash Equilibrium (NE) concept for generating the optimal assignment of tasks and paths, and the attractor selection model (ASM) is introduced to flexibly allocate tasks with consideration of different energy-saving requirements. With comprehensive simulation experiments, we show that the proposed method can save approximately 85% energy compared with the method without task assignment, and it is over 9% and 16% better than the shortest path assignment method and the minimum average delay method, respectively. Besides, the power consumption in AVs with low remaining electricity can be further reduced by over 43%.

Image registration is a basic task in computer vision, for its wide potential applications in image stitching, stereo vision, motion estimation, and etc. Most current methods achieve image registration by estimating a global homography matrix between candidate images with point-feature-based matching or direct prediction. However, as real-world 3D scenes have point-variant photograph distances (depth), a unified homography matrix is not sufficient to depict the specific pixel-wise relations between two images. Some researchers try to alleviate this problem by predicting multiple homography matrixes for different patches or segmentation areas in images; in this letter, we tackle this problem with further refinement, i.e. matching images with pixel-wise, depth-aware homography estimation. Firstly, we construct an efficient convolutional network, theDPH-Net, to predict the essential parameters causing image deviation, the rotation ($R$) and translation ($T$) of cameras. Then, we feed-in an image depth map for the calculation of initial pixel-wise homography matrixes, which are refined with an online optimization scheme. Finally, with the estimated pixel-specific homography parameters, pixel correspondences between candidate images can be easily computed for registration. Compared with state-of-the-art image registration algorithms, the proposedDPH-Nethas the highest performance of 0.912 EPE and 0.977 SSIM, demonstrating the effectiveness of adding depth information and estimating pixel-wise homography into the image registration process.

Challenging due to the absence of color information and low resolution, infrared pedestrian tracking is still of great importance in video surveillance for its ability to distinguish and track persons in night vision. Recent researches show that the CF (Correlation Filters) based methods have a satisfying performance on the infrared pedestrian tracking task when combined with deep features. However, most of these algorithms only rely on the general convolutional features and online model updating to trace object variation, resulting in the lost track of pedestrians with large deformation and infrequent poses. In this paper, we propose a long-term inter-frame graph memory feature as a stronger descriptor for infrared pedestrians. We show that this cross-frame memory graph can enrich deep features with time-variant pedestrian appearance information, boosting the robustness and accuracy of the baseline CF tracker. Extensive experiments and analyses are carried out on the widely-used PTB-TIR dataset, demonstrating the effectiveness and universality of the proposed graph memory features for infrared pedestrians.

Video object segmentation has attracted considerable research interest these years. Top-performing video object segmentation methods mainly rely on fully convolutional neural networks which are specifically trained for predicting high-performance masks, resulting in a lack of preciseness in boundary details. This paper tackles the problem of predicting both mask-accurate and boundary-precise segmentation masks in videos. To solve this problem, we propose a simple and efficient network structure: the Mask-boundAry-Consistent Network (MAC-Net). TheMAC-Netis an end-to-end fully convolutional network, where both mask and boundaries are jointly optimized during training, enabling it to predict masks along with accurate boundaries. An inner-net boundary-computing module is incorporated in theMAC-Netfor producing spontaneously mask-consistent boundaries. We analyze the influence of parameter settings, network constructions of theMAC-Net, and compare with state-of-the-art algorithms on three widely-adopted datasets. Experimental results show that theMAC-Netachieves state-of-the-art performance, demonstrating the effectiveness of its mask-boundary-consistent network structure. We also propose that the boundary module inMAC-Nethas high compatibility, and can be easily adapted to other segmentation-related techniques.

Facial expression recognition (FER) is a key factor in human behavior analysis. Most algorithms deal with FER as a pure classification problem, assuming that expressions are exclusive to each other. In this letter, the problem of FER is tackled from a more detailed view: learning to discriminate expressions with consideration of the secondary information. We propose the Secondary Information aware Facial Expression Network (SIFE-Net) to explore the latent components without auxiliary labeling, and we propose a novel dynamic weighting strategy to teach the SIFE-Net. In contrast to traditional classifiers trained with one-hot labels, the proposed SIFE-Net takes advantage of secondary expression information and has more rational feature distributions. We carry out extensive experiments and analysis on three widely-used FER datasets, i.e. the CK+ dataset, the JAFFE dataset, and the RAF dataset. Experimental results show that the SIFE-Net achieves state-of-the-art performance on all three datasets, which demonstrates the effectiveness of our method.

Online video object segmentation is a challenging task as it entails to process the image sequence timely and accurately. To segment a target object through the video, numerous CNN-based methods have been developed by heavily finetuning on the object mask in the first frame, which is time-consuming for online applications. In this paper, we propose a fast and accurate video object segmentation algorithm that can immediately start the segmentation process once receiving the images. We first utilize a part-based tracking method to deal with challenging factors such as large deformation, occlusion, and cluttered background. Based on the tracked bounding boxes of parts, we construct a region-of-interest segmentation network to generate part masks. Finally, a similarity-based scoring function is adopted to refine these object parts by comparing them to the visual information in the first frame. Our method performs favorably against state-of-the-art algorithms in accuracy on the DAVIS benchmark dataset, while achieving much faster runtime performance.

This paper proposes an end-to-end trainable network, SegFlow, for simultaneously predicting pixel-wise object segmentation and optical flow in videos. The proposed SegFlow has two branches where useful information of object segmentation and optical flow is propagated bidirectionally in a unified framework. The segmentation branch is based on a fully convolutional network, which has been proved effective in image segmentation task, and the optical flow branch takes advantage of the FlowNet model. The unified framework is trained iteratively offline to learn a generic notion, and fine-tuned online for specific objects. Extensive experiments on both the video object segmentation and optical flow datasets demonstrate that introducing optical flow improves the performance of segmentation and vice versa, against the state-of-the-art algorithms.