How beneficial would it be for scientists to have access to an extensive material science database? Imagine the ease of searching for a specific, complex material and instantly obtaining all necessary information, including its physical and chemical properties. Since these properties are often detailed within journal papers, primarily in table formats, the key lies in efficiently extracting these tables from material science journals and integrating them into the database. As an initial measure, a plug-and-playable module was developed that could be plugged into object detectors to extract tables from PDF-formatted material science papers. The biggest challenge was to make the object detectors trained on images to adapt to material science journal papers.

Hazardous gases pose a significant threat, particularly because many of them are imperceptible to humans. Imagine the benefits of an artificial nose capable of electrically detecting these gases, a technology that could be lifesaving in industrial settings where such gases are frequently produced. Current systems, however, face limitations both in physical design and software capabilities. Known as "electronic nose" sensors, these devices are constrained by their operational conditions, and the methods for effectively interpreting their signals are not thoroughly explored. Addressing these challenges, we developed a comprehensive End-to-End Electronic Nose system. Our innovation lies in using Transition Metal Dichalcogenides (TMD) to create a nano-sensor that functions under a broader range of real-world conditions. Leveraging this advanced sensor, we collected extensive gas data and constructed machine learning models specifically for signal classification. This ensures that our system is not just a sensor but a fully functional "electronic nose" capable of both detecting and classifying hazardous gases accurately.