This paper introduces a lightweight deep learning framework designed for real-time defect detection in industrial applications, with a focus on supporting small and medium-sized enterprises (SMEs) facing resource constraints. The proposed system leverages advanced computer vision and deep learning techniques to enable efficient and cost-effective quality control on production lines. The methodology includes a structured pipeline for data processing, model training, and validation, optimized for low-resource environments. The system builds on innovative approaches such as MobileNet to enhance performance while maintaining accessibility. Experimental results demonstrate the framework’s effectiveness, offering a scalable solution for SMEs. The study addresses the research question of whether a low-cost system can provide reliable part validation, achieving objectives of real-time processing, accurate defect detection of bottles and cans to determine if they are defective for example, exhibiting visual defects like dents, scratches, cracks, or contamination spots even under deliberate variations in lighting, orientation, and surface reflections. Limitations and future enhancements are also discussed to guide ongoing development in industrial automation.