Detecting small objects in images poses significant challenges due to their limited pixel representation and the difficulty in extracting sufficient features, often leading to missed or false detections. To address these challenges and enhance detection accuracy, this paper presents an improved small object detection algorithm, CRL-YOLOv5. The proposed approach integrates the Convolutional Block Attention Module (CBAM) attention mechanism into the C3 module of the backbone network, which enhances the localization accuracy of small objects. Additionally, the Receptive Field Block (RFB) module is introduced to expand the model’s receptive field, thereby fully leveraging contextual information. Furthermore, the network architecture is restructured to include an additional detection layer specifically for small objects, allowing for deeper feature extraction from shallow layers. When tested on the VisDrone2019 small object dataset, CRL-YOLOv5 achieved an mAP50 of 39.2%, representing a 5.4% improvement over the original YOLOv5, effectively boosting the detection precision for small objects in images.
19 November 2025 in Biomedical Physics & Engineering Express
To enable efficient and accurate retinal lesion segmentation on resource-constrained point-of-care Optical Coherence Tomography (OCT) systems, we propose OCTSeg-UNeXt, an ultralight hybrid Convolution-Multilayer Perceptron (Conv-MLP) network optimized for OCT image analysis. Built upon the UNeXt architecture, our model integrates a Depthwise-Augmented Scale Context (DASC) module for adaptive multi-scale feature aggregation, and a Group Fusion Bridge (GFB) to enhance information interaction between the encoder and decoder. Additionally, we employ a deep supervision strategy during training to improve structural learning and accelerate convergence. We evaluated our model using three publicly available OCT datasets. The results of the comparative experiments and ablation experiments show that our method achieves powerful performance in multiple key indicators. Importantly, our method achieves this high performance with only 0.187 million parameters (Params) and 0.053 G Floating-Point Operations Per second (FLOPs), which is significantly lower than UNeXt (0.246M, 0.086G) and UNet (17M, 30.8G). These findings demonstrate the proposed method’s strong potential for deployment in Point-of-Care Imaging (POCI) systems, where computational efficiency and model compactness are crucial.