Semi-Supervised Domain Adaptation for Wildfire Detection

TL;DR

This work tackles wildfire detection under domain shift with limited labeled target data by introducing semi-supervised domain adaptation for object detection. It presents LADA, a Location-Aware SSDA framework that combines CoordConv-inspired location awareness with a teacher-student architecture, pseudo-labeling, background-image augmentation, and adversarial domain alignment. A new HPWREN-based dataset and labeling policy expands label diversity by about 30x and defines SSDA protocols at $0.5\%$, $1.0\%$, and $3.0\%$ target labels, achieving a notable $3.8\%$ improvement in mean Average Precision with only $1\%$ target labels. The work provides a practical benchmark and robust baseline for wildfire detection across domains, with potential impact for both academic research and industry deployment.

Abstract

Recently, both the frequency and intensity of wildfires have increased worldwide, primarily due to climate change. In this paper, we propose a novel protocol for wildfire detection, leveraging semi-supervised Domain Adaptation for object detection, accompanied by a corresponding dataset designed for use by both academics and industries. Our dataset encompasses 30 times more diverse labeled scenes for the current largest benchmark wildfire dataset, HPWREN, and introduces a new labeling policy for wildfire detection. Inspired by CoordConv, we propose a robust baseline, Location-Aware Object Detection for Semi-Supervised Domain Adaptation (LADA), utilizing a teacher-student based framework capable of extracting translational variance features characteristic of wildfires. With only using 1% target domain labeled data, our framework significantly outperforms our source-only baseline by a notable margin of 3.8% in mean Average Precision on the HPWREN wildfire dataset. Our dataset is available at https://github.com/BloomBerry/LADA.

Figures (5)

• Figure 1: semi-supervised Domain Adaptation
• Figure 2: Original HPWREN labeled image (Left) vs. Proposed labeled image (Right)
• Figure 3: Location Aware semi-supervised Domain Adaptation Network. We omitted the second stage regression and classification heads for simplicity.
• Figure 4: Overall Diagram of our training process. We also use background images for training.
• Figure 5: Example of our labeled dataset