In real-life scenarios, the accuracy of person re-identification (Re-ID) is subject to the limitation of camera hardware conditions and the change of image resolution caused by factors such as camera focusing errors. People call this problem cross-resolution person Re-ID. In this paper, we improve the recognition accuracy of cross-resolution person Re-ID by enhancing the image enhancement network and feature extraction network. Specifically, we treat cross-resolution person Re-ID as a two-stage task: the first stage is the image enhancement stage, and we propose a Super-Resolution Dual-Stream Feature Fusion sub-network, named SR-DSFF, which contains SR module and DSFF module. The SR-DSFF utilizes the SR module recovers the resolution of the low-resolution (LR) images and then obtains the feature maps of the LR images and super-resolution (SR) images, respectively, through the dual-stream feature fusion with learned weights extracts and fuses feature maps from LR and SR images in the DSFF module. At the end of SR-DSFF, we set a transposed convolution to visualize the feature maps into images. The second stage is the feature acquisition stage. We design a global-local feature extraction network guided by human pose estimation, named FENet-ReID. The FENet-ReID obtains the final features through multistage feature extraction and multiscale feature fusion for the Re-ID task. The two stages complement each other, making the final pedestrian feature representation has the advantage of accurate identification compared with other methods. Experimental results show that our method improves significantly compared with some state-of-the-art methods.
Knowledge Distillation has shown promising results for classifying skin lesions in dermatology images. Traditional knowledge distillation typically involves the student model passively mimicking the teacher model's knowledge. We propose utilizing AdaBoost to enable the student to actively mine the teacher's learning representation for skin lesion classification. This paradigm allows the student to determine the “granularity” in mining the teacher's knowledge. As the student's learning process progresses, it can become challenging to pinpoint specific learning difficulties, especially with potential interference from the teacher. To address this issue, we introduce a variational difficulty mining strategy to reduce the impact of such interference. This strategy involves the distillation module capturing more nuanced classification difficulties by extracting information from the node's $$l$$ th hops. By maximizing the mutual information between the teacher and student, we effectively filter out noise interference from these nuanced difficulties. Our proposed framework, Variational AdaBoost Knowledge Distillation (VAdaKD), allows the student to actively mine and leverage the teacher's knowledge for improved skin lesion classification. Our proposed method performs satisfactorily on three benchmark datasets: the Dermnet dataset, ISIC 2019 dataset, and HAM10000 dataset, respectively. Specifically, our method shows an improvement of 2–3% over the baseline on the Dermnet dataset and outperforms the best results of the other compared methods by 1%. Experimental results and visualization performance indicate that our proposed method effectively captures the learning difficulties and achieves better visualized t-distributed stochastic neighbor embedding classification results. Our code is available at https://github.com/25brilliant/VAdaKD.
Illumination significantly affects lane detection systems. Existing lane detection models perform well under well-lit conditions but exhibit issues such as reduced detection accuracy and high false-negative rates in low-light scenarios. To address these issues, we propose an Instance Feature Association Network (IFANet) tailored for lane detection under low-light conditions, incorporating the prior knowledge of lane lines being elongated and continuous. We design a Multi-Scale Feature Enhancer (MSFE) that utilizes convolutional kernels of different scales to capture detailed features. Deformable convolutions are employed to generate offsets, dynamically adjusting the positions of the convolutional kernels to better capture the details and shape variations of lane lines. A lane instance separation module is introduced, which uses heatmaps to locate the starting points of lanes and combines global information to generate corresponding masks for each lane, thereby separating lane instances. To further enhance feature sensitivity under low-light conditions, we design a Multi-Head Attention Instance Association module (MAIA) to facilitate information exchange between adjacent lane line instances. Finally, lane line fitting is achieved by locating key points and calculating offsets. IFANet achieves F1 scores of 75.68% and 72.03% in the CULane dataset and night-time scenarios, respectively, significantly outperforming other models of its kind, demonstrating the effectiveness of our model in detecting lane lines under low-light conditions.
4 September 2026 in ACM Transactions on Computing for Healthcare
Accurate fracture detection in medical imaging is pivotal for intelligent orthopedic diagnostic systems, yet deploying high-capacity detection models on resource-constrained platforms remains a critical cyber-physical challenge. Conventional feature distillation based on Mean Squared Error (MSE) performs point-wise feature regression but does not explicitly model higher-order statistical relationships, which may lead to over-smoothed responses and weaken the preservation of subtle fracture structures. To overcome this systemic limitation, we propose Structured Feature Decoupling Knowledge Distillation (SFD-KD). This framework decomposes teacher features into linearly combined first-order semantics and multi-order statistics, bypassing computationally expensive covariance modeling. Concretely, SFD-KD integrates three specialized modules: the Multi-Order Statistic Extractor (MOSE) for multi-order structural alignment, First-Order Statistic Extractor (FOSE)-SVD for teacher-side semantic extraction via Singular Value Decomposition (SVD), and FOSE-FFT for student-side semantic stabilization using Fast Fourier Transform (FFT) with a Learnable Spectral Filter (LSF). Extensive experiments on the GRAZPEDWRI-DX and FracAtlas datasets demonstrate that SFD-KD consistently outperforms vanilla KD by +1.4~+4.4% mAP@0.5:0.95 across both Faster R-CNN and YOLOv8 frameworks. Ablation studies confirm the efficacy of the proposed statistical decoupling, while visualizations reveal superior preservation of multi-order fracture patterns. Notably, SFD-KD enables robust deployment in clinical cyber-physical systems with negligible inference overhead. Our code is available at: https://github.com/6720230811/SFD-KD.
Aiming at the problem of incomplete intra-class change coverage and counting error caused by insufficient samples in few-shot object counting, this paper proposes a few-shot object counting network combining self-support matching and attention mechanism. The network uses a deep cross-correlation module (DW-Corr) to support features as convolution kernels to achieve accurate target tracking. At the same time, the self-supporting prototype generation module (GSSP) is used to generate self-supporting prototypes to reduce the appearance differences between samples. Finally, the cross-attention mechanism (CA) is used to fuse the self-support prototype and support features generated by query features. Experiments on FSC-147 and CAPRK datasets show that the network is superior to other methods in few-shot object counting, and has strong competitiveness and performance.