Learning to Calibrate for Reliable Visual Fire Detection
This work addresses the problem of reliable fire detection for those relying on computer vision systems, particularly in applications where early detection is crucial for human safety and property protection, and presents an incremental improvement.
The authors tackled the problem of overconfidence in deep learning models for visual fire detection, achieving more reliable results by incorporating uncertainty modeling, with experiments on two datasets. The proposed method improves the balance between classification accuracy and reliable decision-making.
Fire is characterized by its sudden onset and destructive power, making early fire detection crucial for ensuring human safety and protecting property. With the advancement of deep learning, the application of computer vision in fire detection has significantly improved. However, deep learning models often exhibit a tendency toward overconfidence, and most existing works focus primarily on enhancing classification performance, with limited attention given to uncertainty modeling. To address this issue, we propose transforming the Expected Calibration Error (ECE), a metric for measuring uncertainty, into a differentiable ECE loss function. This loss is then combined with the cross-entropy loss to guide the training process of multi-class fire detection models. Additionally, to achieve a good balance between classification accuracy and reliable decision, we introduce a curriculum learning-based approach that dynamically adjusts the weight of the ECE loss during training. Extensive experiments are conducted on two widely used multi-class fire detection datasets, DFAN and EdgeFireSmoke, validating the effectiveness of our uncertainty modeling method.