MTL-FFDET: A Multi-Task Learning-Based Model for Forest Fire Detection

Abstract

Deep learning-based forest fire vision monitoring methods have developed rapidly and are becoming mainstream. The existing methods, however, are based on enormous amounts of data, and have issues with weak feature extraction, poor small target recognition and many missed and false detections in complex forest scenes. In order to solve these problems, we proposed a multi-task learning-based forest fire detection model (MTL-FFDet), which contains three tasks (the detection task, the segmentation task and the classification task) and shares the feature extraction module. In addition, to improve detection accuracy and decrease missed and false detections, we proposed the joint multi-task non-maximum suppression (NMS) processing algorithm that fully utilizes the advantages of each task. Furthermore, considering the objective fact that divided flame targets in an image are still flame targets, our proposed data augmentation strategy of a diagonal swap of random origin is a good remedy for the poor detection effect caused by small fire targets. Experiments showed that our model outperforms YOLOv5-s in terms of mAP (mean average precision) by 3.2%, APs (average precision for small objects) by 4.8%, ARs (average recall for small objects) by 4.0%, and other metrics by 1% to 2%. Finally, the visualization analysis showed that our multi-task model can focus on the target region better than the single-task model during feature extraction, with superior extraction ability.

A multitask model for realtime fish detection and segmentation based on YOLOv5

“…Different from traditional single-task learning model, which seeks to use a specific model ( Ma et al, 2018 ) to accomplish the task, multi-task learning model is more effective to improve the generalization of models. In addition, computational repetition can be minimized, inference speed can be increased, and memory utilization can be reduced by sharing layers across multiple tasks ( Lu et al, 2022 ).…”

Deforestation detection using deep learning-based semantic segmentation techniques: a systematic review

“…Similarly, Guan et al (2022) extracted ground truth for each image using LabelMe software tools from the Flame dataset. Lu et al (2022) employed polygon annotation with LabelMe software tools to outline the flame target. However, GIS software tools, as demonstrated in the works by Chiang et al (2020), were utilized for the manual annotation of datasets in UAV imagery.…”

“…It is worth noting that conventional machine learning algorithms, such as Early Fusion (EF) (de Andrade et al, 2022), Multi-Task Learning (MTL) (Lu et al, 2022), and Random Forest (RF) (Dal Molin and Rizzoli, 2022), continue to be popular choices among researchers in the context of image segmentation, including for deforestation detection applications. These algorithms have been employed in 8% of the studies included in the final selection.…”

Dual-Dataset Deep Learning for Improved Forest Fire Detection: A Novel Hierarchical Domain-Adaptive Learning Approach

“…MTL models trained to concurrently execute multiple tasks offer a nuanced understanding of fire dynamics. This approach has the potential to mitigate the tendency of models to become overly specialized, a notable challenge when training is confined to controlled datasets [22]. Recent developments in this field are exemplified in the study detailed in Reference [15].…”