Traditional warehouse management heavily relies on manual processes such as data entry and barcoding, which are time-consuming, labor-intensive, and susceptible to errors. While automated solutions like RFID improve efficiency, their high implementation costs make them inaccessible for many small and medium-sized enterprises (SMEs). Additionally, accurately tracking visually similar objects, such as identical pallets, remains a significant challenge, particularly in dynamic and cluttered warehouse environments where maintaining object identities across frames is crucial. This study proposes a cost-effective warehouse tracking system that leverages deep learning and computer vision technologies to address these challenges. The system utilizes YOLOv8 for precise object detection and a custom Residual CBAM-based network for enhanced feature extraction, enabling real-time pallet monitoring. DeepSORT is employed to ensure robust object tracking by minimizing ID switches, even among visually similar items, while a multi-angle camera setup enhances coverage and tracking accuracy. By integrating these advanced techniques, the proposed system facilitates warehouse digitalization, enabling real-time object tracking, historical movement analysis, and optimized logistics workflows. Unlike traditional high-cost solutions, this approach offers a scalable and efficient alternative, improving operational efficiency and workforce management while making automation more accessible to SMEs.