CS426 - Senior Project - Spring 2023
Computer Science and Engineering (CSE)
University of Nevada Reno (UNR)
     Deep-learning (DL) based object detection algorithms are beneficial to areas of climate science involving wildfire prevention, observation, and intelligence. As wildfires become more prevalent, the need for advanced techniques in processing video feeds has become crucial for smoke detection. Existing DL-based techniques have been developed based on convolution networks and are proven to be effective in wildfire detection. However, these datasets are primarily commercial and/or closed-sourced. One of the few open-source models is Nemo, a benchmark for fine-grained wildfire smoke detection in the incipient stage of a wildfire. Using this model as a basis, our team intends to collaborate with its authors to build a convenient user friendly interface.
     Our senior project plans to accomplish three main goals with integrating the Nemo smoke detection model into a web application. The web application will help bring exposure to the project through an attractive and simple to use interface, in the hopes of building recognition of this model within the field of fire science. The user-friendly design is built around the goal of making the model more accessible to users who may not be familiar with computer science skills and technology. Finally, the model aims to provide a simple way for users to contribute to the project by labeling and submitting images to help improve the smoke detection model currently in place. The intended users of this web application range from ordinary people interested in analyzing the risk of a wildfire in its incipient stage (based on images and videos) with proximity to their businesses or properties, users who are interested in analyzing images and videos of wildfires, emergency service members such as firefighters or forestry personnel with the goal of rapidly analyzing the potential risk of a fire, and researchers in the field who want to further expand or contribute their knowledge to building and improving the current model and software.