In light of the increasing application of deep learning across various domains and the advancing computational power of computers, this study presents a novel approach using transfer learning-based deep convolutional neural network algorithm for in-situ biodeteriogen detection and classification on paper. This development aims to enhance the accessibility of biodeteriogen identification tools and support more efficient work in the fields of archaeology and conservation. Four of the most commonly identified biodeteriogens on paper-based relics were selected for investigation. We demonstrated the feasibility of non-invasive diagnosis of biocontaminated relics using deep learning model during the testing phase with 90.29% accuracy and high precision. This approach holds significant potential for assisting conservation decision-making processes. By leveraging the power of deep learning and computer vision, this research opens new avenues for effective and timely identification of microorganism contaminants, enabling proactive conservation strategies for paper-based artifacts.