Fire and Smoke Detection with Burning Intensity Representation

TL;DR

A new Attentive Fire and Smoke Detection Model (a-FSDM) is proposed, which retains the robust feature extraction and fusion capabilities of conventional detection algorithms but also redesigns the detection head specifically for transparent targets in FSD, termed the Attentive Transparency Detection Head (ATDH).

Abstract

An effective Fire and Smoke Detection (FSD) and analysis system is of paramount importance due to the destructive potential of fire disasters. However, many existing FSD methods directly employ generic object detection techniques without considering the transparency of fire and smoke, which leads to imprecise localization and reduces detection performance. To address this issue, a new Attentive Fire and Smoke Detection Model (a-FSDM) is proposed. This model not only retains the robust feature extraction and fusion capabilities of conventional detection algorithms but also redesigns the detection head specifically for transparent targets in FSD, termed the Attentive Transparency Detection Head (ATDH). In addition, Burning Intensity (BI) is introduced as a pivotal feature for fire-related downstream risk assessments in traditional FSD methodologies. Extensive experiments on multiple FSD datasets showcase the effectiveness and versatility of the proposed FSD model. The project is available at \href{https://xiaoyihan6.github.io/FSD/}{https://xiaoyihan6.github.io/FSD/}.

Figures (7)

• Figure 1: The framework of a-FSDM. It retains three parts: the feature extraction and fusion networks of the general object detection algorithm, and a novel ATDH.
• Figure 2: Baseline model comparison across different datasets. Fire, Smoke and $\mathrm{mAP}$ are given in the subsetion “Setting and Details”. “avg” represents the average of mAP (mean Average Precision) values of all models across the FSD datasets. “s" represents the input image of small size, while “l" represents the input image of large size. F-RCNN means Faster RCNN.
• Figure 3: Comparison between generic detection heads and the Attention Transparency Detection Head (ATDH) across the MS-FSDB. Fire, Smoke and $\mathrm{mAP}$ are given in the subsetion “Setting and Details”. “s" represents the input image of small size, while “l" represents the input image of large size.
• Figure 4: The performance of different models. It shows in detail the performance of various seven detection models in terms of key metrics such as detection accuracy, $\mathrm{mAP}$, Performance (Gflops) and number of parameters (M) for each category, where Fire, Smoke and $\mathrm{mAP}$ are given in the subsetion “Setting and Details”. And, we use the input image of small size used the miniMS-FSDB. In models, 26 represents Yolov5, 47 stands for Yolov8, 18 denotes SSD, 45 signifies RetinaNet, 16 means Faster RCNN, 36 refers to FCOS, and ours represents the proposed a-FSDM model. Additionally, s, x, n, l, m represent different versions of the model.
• Figure 5: An image of containing transparent flame, (a) an enlarged patch of Transparent Background (flame) and Foreground (sofa). (b) Detection Result of the state-of-the-art generic detector 36. The blue box in (b) represents ground truth, and the phenomenon of missed detection occurs in (b).