Fire and Smoke Detection with Burning Intensity Representation
Xiaoyi Han,
Yanfei Wu,
Nan Pu,
Zunlei Feng,
Qifei Zhang,
Yijun Bei,
Lechao Cheng
Zhejiang University & University of Trento & Hefei University of Technology
Accepted to MM Asia 2024
Dear Visitors,
We would like to inform you that the currently provided code supports only the following object detection models (or other components):
- SSD
- RetinaNet
- FCOS
- Attentive Transparency Detection Head (ATDH) [We placed the ATDH in FCOS]
Best regards,
Xiaoyi Han
python == 3.8.5
torch == 1.11.0+cu113
torchaudio == 0.11.0+cu113
torchvision == 0.12.0+cu113
pycocotools == 2.0.4
numpy
Cython
matplotlib
opencv-python (maybe you want to use skimage or PIL etc...)
scikit-image
tensorboard
tqdm
...
I use Ubuntu20.04 (OS).
# Project
FSDmethod path: /data/PycharmProject/FSDmethod
├── assets
├── README.md
├── SSD ( layout-> the same as FCOS)
├── RetinaNet ( layout-> the same as FCOS)
└── MyFireNet (FCOS)
├── checkpoints
├── configs
├── data
├── log (accuracy)
├── models (Head->ATDH)
├── options
├── results (visualization)
├── tensorboard
├── tools
└── utils
# Dataset
Dataset path: /data/
├── 1_VisiFire (layout -> the same as MS-FSDB)
├── 2_FIRESENSE (layout -> the same as MS-FSDB)
├── 3_furg_fire-dataset (layout -> the same as MS-FSDB)
├── 4_BoWFireDataset (layout -> the same as MS-FSDB)
├── 5_FIRE_SMOKE_DATASET (layout -> the same as MS-FSDB)
└── MS-FSDB
├──data
├──images
├──labels
└──layout
We use MyFireNet instead of the name FCOS.
# Object Detection(SSD, RetinaNet, FCOS)
# path:/data/PycharmProject/FSDmethod/Object Detection/tools
# Training
run train.py
# Evaluation
run eval_voc.py
# Visualization
run visualize.py
Table 1: Baseline model comparison across different datasets. Fire, Smoke and mAP are given in the subsection "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.
Table 2: Comparison between generic detection heads and the Attention Transparency Detection Head (ATDH) across the MS-FSDB. Fire, Smoke and "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.
Table 3: The attention mechanism algorithm added to the baseline (FCOS) on the MS-FSDB. Fire, Smoke and "mAP" are given in the subsetion “Setting and Details”. the input image of small size is used.
The Detection of Transparent Targets Images in FSD, (a) the false results of generic detection, (b) that the proposed method successfully detected the previous failure result. In the diagram, blue boxes represent ground truth and red boxes represent predicted results.
@inproceedings{han2024mmAsia,
author = {Han, Xiaoyi and Wu, Yanfei and Pu, Nan and Feng, Zunlei and Zhang, Qifei and Bei, Yijun and Cheng, Lechao},
title = {Fire and Smoke Detection with Burning Intensity Representation},
year = {2024},
isbn = {9798400712739},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3696409.3700165},
doi = {10.1145/3696409.3700165},
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 https://xiaoyihan6.github.io/FSD/.1},
booktitle = {Proceedings of the 6th ACM International Conference on Multimedia in Asia},
articleno = {5},
numpages = {8},
keywords = {Fire and Smoke Detection, Attentive Transparency Detection Head, Burning Intensity},
location = {
},
series = {MMAsia '24}
}
Note:Could you please give me a "one-click triple support"🔥 ("Star"🚀,"Fork"🔖,"Issues"❓)