<p>Small object detection poses substantial challenges due to limited pixel count and sparse features, yet it holds immense significance in applications like autonomous driving and unmanned aerial vehicles. This paper introduces an algorithm derived from YOLOv10, significantly enhancing the accuracy of detecting small objects. We have devised a novel module, the Context-aware and Enhanced Capture Module (C2A), which addresses the intricacies of small object detection by skillfully integrating multi-scale features and contextual information to bolster recognition and capture capabilities. Additionally, we incorporate the Receptive Field Attention Convolution (RFAConv) mechanism, leveraging attention weights to precisely evaluate the significance of each receptive field position’s information, facilitating the extraction of crucial small object features. Furthermore, we introduce Adaptive Convolution Kernel (AKConv) technology, which dynamically adjusts the convolution kernel to effectively handle small objects of varying sizes and shapes, thereby enhancing detection performance. Evaluations on the VisDrone2021 dataset reveal that our algorithm achieves AP50 and mAP scores of 29.3 and 16.4. This substantial performance uplift underscores the efficacy of our proposed algorithm in small object detection tasks.</p>