Object detection, a vital aspect of computer vision, predominantly relies on convolutional neural networks (CNNs) in recent advanced methods. Researchers have persistently formulated diverse architectures to augment the performance of these networks. Convolutional block attention module stands out as a significant attention mechanism for CNNs incorporating channel and spatial attention units to capture detailed feature information from input images. However, in conventional convolutional processes, CBAM uses a fixed kernel size, which may pose a notable limitation to learning variances in the input data. In this study, we present an Adaptive Convolutional Block Attention Module (ACBAM), a novel extension of CBAM by introducing adaptive kernel sizes, leveraging learning parameters (K) to dynamically adjust the filter sizes during convolutional operations. We integrated the proposed module into the YOLOv8 framework and evaluated the modified network using KITTI and Pascal VOC datasets. The results from experiments demonstrate that, YOLOv8 with the proposed attention mechanism, reach an mAP of 96.90% for KITTI and 94.67% for PASCAL VOC which is a significant improvement over the original YOLOv8 model.