Furniture plates, being a crucial raw material in furniture manufacturing, often exhibit various defects during their production. These defects potentially compromise the quality of the finished furniture products and inflate production costs. Traditional methods for detecting plate defects face challenges, particularly in identifying less distinct features and handling surface noise, leading to suboptimal detection results. To address these limitations, this study introduces a specialized dataset named the “Furniture Plate Defect Dataset” for evaluating and improving defect detection algorithms more comprehensively. Furthermore, the study employs an enhanced version of the YOLOv5 algorithm, augmented with a small object detection head and incorporated with a Convolutional Block Attention Module (CBAM) to specifically optimize for plate defects. Experimental results demonstrate that with extensive training and fine-tuning on the newly constructed dataset, the enhanced YOLOv5 algorithm exhibits significant improvements in defect detection in furniture plates. The upgraded algorithm is adept at accurately identifying both texture-related and shape-related defects thereby substantially improving the detection’s accuracy and robustness. In summary, the refined YOLOv5 algorithm excels in defect detection, reaching an mAP50 of 81.6%, indicating its considerable potential for application.