YOLC is an efficient and effective framework for detecting small objects in aerial images. It builds upon the anchor-free object detector CenterNet and introduces a Local Scale Module (LSM) to adaptively search for cluster regions, enabling accurate detection. The regression loss is modified using Gaussian Wasserstein distance (GWD) to improve bounding box quality, and deformable convolutions are used to enhance small object detection. The framework also includes a refined detection head with a disentangled heatmap branch for precise localization of different object categories. Extensive experiments on the VisDrone2019 and UAVDT datasets demonstrate that YOLC outperforms existing methods in terms of accuracy and efficiency. The LSM is an unsupervised module that simplifies training and reduces complexity, while the GWD-based loss improves small object detection. The improved detection head with deformable convolutions and disentangled heatmap branch further enhances performance. YOLC achieves high accuracy and efficiency in detecting small objects in large-scale aerial images.YOLC is an efficient and effective framework for detecting small objects in aerial images. It builds upon the anchor-free object detector CenterNet and introduces a Local Scale Module (LSM) to adaptively search for cluster regions, enabling accurate detection. The regression loss is modified using Gaussian Wasserstein distance (GWD) to improve bounding box quality, and deformable convolutions are used to enhance small object detection. The framework also includes a refined detection head with a disentangled heatmap branch for precise localization of different object categories. Extensive experiments on the VisDrone2019 and UAVDT datasets demonstrate that YOLC outperforms existing methods in terms of accuracy and efficiency. The LSM is an unsupervised module that simplifies training and reduces complexity, while the GWD-based loss improves small object detection. The improved detection head with deformable convolutions and disentangled heatmap branch further enhances performance. YOLC achieves high accuracy and efficiency in detecting small objects in large-scale aerial images.