Xue-Feng Zhao

HFQI-YOLO: a lightweight and efficient underwater object detection algorithm

Underwater object detection plays a crucial role in fisheries resource assessment and ecological environment protection. Current underwater object detection models are characterized by large parameter sizes and high computational costs, which hinder the simultaneous achievement of lightweight deployment and high detection accuracy for real-time and resource-limited underwater platforms. Therefore, this paper proposes a lightweight underwater object detection method named HFQI-YOLO based on YOLO11n. First, a High-level Screening Path Aggregation Network with the Dysample is designed to replace the traditional PANet, which effectively improves multi-scale feature fusion quality while significantly reducing computational resources through bidirectional feature screening and dynamic upsampling. Subsequently, the Feature Complementary Mapping module is embedded into the backbone to mitigate the mismatch between deep semantic representations and shallow spatial details via effective semantic spatial complementary interactions. Moreover, a quality-aware shared detection head is introduced to enhance detection reliability. This design simultaneously decreases model parameters and resolves the inconsistency between classification confidence and localization precision. Finally, Inner WIoU is employed as the loss function to refine bounding box regression and increase sensitivity to small object instances. Experimental results demonstrate that the proposed algorithm outperforms the YOLO11n baseline on the URPC2020 dataset, achieving only 1.5 M and 4.3 GFLOPs, which are reduced by 41.7% and 31.7% respectively, and an increase of 0.9 percentage points in mAP@0.5%–82.6%. Furthermore, the generalization ability and robustness of the algorithm are validated on the RUOD dataset, further demonstrating its superior performance.

Reliability-aware disentangled adaptive network for multimodal sentiment analysis

Compact Occlusion-Robust Facial Expression Recognition via Clean-Anchored Hard Occlusion Fine-Tuning

Facial occlusion removes expression-relevant evidence and remains a major source of error in camera-based affective sensing. Existing approaches to occlusion-robust facial expression recognition often rely on specialized attention, reconstruction, semantic, or geometric pipelines, whereas aggressive training using synthetic occlusions may impair discrimination on clean images or overfit to synthetic corruption patterns. We address this tension between cleanness and robustness through clean-anchored hard occlusion fine-tuning (CA-HOFT). HardMix samples structured and random occlusion modes according to a facial region-weighted distribution. An explicit classification branch for the non-HardMix source view preserves ground-truth supervision, while a fixed reference teacher initialized from the preceding mixed-occlusion stage supplies a stationary distribution for both paired student views. These signals jointly train a single classifier while retaining a single-backbone inference pathway. Across five independent training runs, the ResNet-18 student achieves 90.08% accuracy on the Real-World Affective Faces Database (RAF-DB) and 87.38% accuracy with 83.34% macro-F1 on Occlusion-RAF-DB, with corresponding sample standard deviations of 0.21, 0.12, and 0.32 percentage points. The retained model has 11.18 million parameters and requires 1.814 giga multiply–accumulate operations (GMACs). Controlled comparisons and ablations indicate that it has the most favorable observed cleanness–robustness trade-off among the tested epoch-matched alternatives; however, fixed-checkpoint comparisons on Occlusion-RAF-DB are not significant after Holm correction. AffectNet-8 and evaluations using natural occlusion provide supporting evidence, while broader cross-domain validation remains future work.