🤖 AI Summary
Vanilla Transformers, optimized for mid-to-high-level semantic modeling, struggle to capture smoke-specific low-level features—such as transparency, texture, and chromatic aberration—limiting wildfire smoke detection performance. To address this, we propose CCPE (Cross-Contrastive Patch Embedding), a backbone that enhances fine-grained feature representation via multi-scale spatial contrastive learning, and SNSM (Separable Negative Sampling Mechanism), which mitigates supervision ambiguity from hard negative samples. Furthermore, we introduce SKLFS-WildFire, the first large-scale, real-world wildfire smoke detection benchmark. Evaluated on FIgLib and SKLFS-WildFire, our method achieves an mAP improvement of up to 8.2% over YOLOv5/v8, Faster R-CNN, and the baseline Swin Transformer. All code and datasets are publicly released.
📝 Abstract
Vanilla Transformers focus on semantic relevance between mid- to high-level features and are not good at extracting smoke features as they overlook subtle changes in low-level features like color, transparency, and texture which are essential for smoke recognition. To address this, we propose the Cross Contrast Patch Embedding (CCPE) module based on the Swin Transformer. This module leverages multi-scale spatial contrast information in both vertical and horizontal directions to enhance the network's discrimination of underlying details. By combining Cross Contrast with Transformer, we exploit the advantages of Transformer in global receptive field and context modeling while compensating for its inability to capture very low-level details, resulting in a more powerful backbone network tailored for smoke recognition tasks. Additionally, we introduce the Separable Negative Sampling Mechanism (SNSM) to address supervision signal confusion during training and release the SKLFS-WildFire Test dataset, the largest real-world wildfire testset to date, for systematic evaluation. Extensive testing and evaluation on the benchmark dataset FIgLib and the SKLFS-WildFire Test dataset show significant performance improvements of the proposed method over baseline detection models. The code and data are available at github.com/WCUSTC/CCPE.
