AUTHOR=Wang Chuansheng , Grau Antoni , Guerra Edmundo , Shen Zhiguo , Hu Jinxing , Fan Haoyi 

TITLE=Semi-supervised wildfire smoke detection based on smoke-aware consistency

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 13 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.980425

DOI=10.3389/fpls.2022.980425

ISSN=1664-462X

ABSTRACT=The semi-transparency property of smoke integrates it highly with background contextual information in the image, which results in great visual differences in different areas. In addition, the limited annotation of smoke images from real forest scenarios brings more training challenges to models. In this paper,  we design a semi-supervised learning strategy, named smoke-aware consistency (SAC), to maintain the pixel and context perceptual consistency in different backgrounds, which can improve the robustness of the model under different scenarios and visually degrades. Additionally, we propose a smoke detection strategy with triple classification assistance to discriminate smoke and smoke-like objects. Finally, we simplified the fire smoke detection network LFNet to LFNet-v2, due to the proposed SAC and triple classification assistance that can accomplish the functions of some specific modules. Extensive experiments validate that the proposed method significantly outperforms state-of-the-art object detection algorithms on wildfire smoke datasets and obtains satisfactory performances in challenging bad weather conditions.