Flame detection using deep learning

Flame detection is an increasingly important issue in intelligent surveillance. In fire flame detection, we need to extract visual features from video frames for training and test. Based on them, a group of shallow learning models have been developed to detect flames, such as color-based model, fuzzy-based model, motion and shape-based model, etc. Deep learning is a novel method which could be much efficient and accurate in flame detection. In this paper, we use YOLO model to implement flame detection and compare it with those shallow learning methods so as to determine the most efficient one for flame detection. Our contribution of this paper is to make use of the optimized YOLO model for flame detection from video frames. We collected the dataset and trained them using Google platform TensorFlow, the obtained accuracy of our proposed flame detection is up to 76%.