Accurate and consistent inspection of surface defects is crucial for ensuring the quality control of footwear products. However, the industry predominantly relies on visual inspections, which are time-consuming and susceptible to subjective variations. This study presents an automated surface inspection system for footwear products utilizing deep learning-based object detection techniques. A multi-view camera system was first installed on a conveyor belt to capture the image data. Subsequently, an open-world object detection model was implemented to detect and center footwear products within the frame. To address more complex defect detection tasks, a closed-set object detection model was trained using annotated footwear images. The optimal object detector was selected through a comparative analysis of various YOLO (You Only Look Once) object detection models, achieving a maximum F1 score of 67%. Finally, the model was deployed using SAHI (Slicing-Aided Hyper Inference) framework for real-time inference. The experimental results demonstrated that the proposed system is able to detect tiny surface defects on footwear products within a second. Consequently, the proposed system can reduce manufacturing costs and enhance quality control processes in the footwear industry.