IDFire: Image Dataset for Indoor Fire Load Recognition

Citation Author(s):
Jia-Rui
Lin
Tsinghua University
Yu-Cheng
Zhou
Tsinghua University
Ke-Xiao
Yan
Tsinghua University
Zhen-Zhong
Hu
Tsinghua University
Submitted by:
Jia-Rui Lin
Last updated:
Mon, 06/20/2022 - 20:01
DOI:
10.21227/qkk3-2145
Data Format:
Link to Paper:
License:
0
0 ratings - Please login to submit your rating.

Abstract 

Accurate fire load (combustible objects) information is crucial for safety design and resilience assessment of buildings. Traditional fire load acquisition methods, such as fire load survey, which are time-consuming, tedious, and error-prone, failed to adapt to dynamic changed indoor scenes. As a starting point of automatic fire load estimation, fast recognition and detection of indoor fire load are important. Thus, A dataset containing images of indoor scenes and annotations of instance segmentation is developed in this research. In total, 1015 images are contained in the dataset, distributed across five typical scenes: bedroom, dining room, hospital, living room, and office. 

Instructions: 

Setup & Usage

·        Install Pytorch 1.6+ and detectron2

·        Clone or download the repo

git clone https://github.com/Zhou-Yucheng/fire-load-detection.gitcd fire-load-detection/src

·        Unzip the dataset trainval1k.zip in data/indoor-scene

·        Run python3 train.py --help for more information about usage

·        Run train.py with arguments, for example:

 

python3 train.py -m R50 -b 4 -l 2e-3 -i 6k --step 4k