Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper addresses the domain shift challenge in street-view weather recognition by proposing Weather-Aware Adversarial Discriminative Domain Adaptation (WA-ADDA), which conditions the domain discriminator on predicted weather to learn domain-invariant and weather-sensitive features. A novel multi-dataset benchmark is introduced, unifying diverse non-street-view sources with real street-view targets and a standardized evaluation protocol. WA-ADDA consistently enhances street-view performance across various backbones, achieving improved per-class recalls in adverse conditions while maintaining clear-weather accuracy.
Adversarial domain adaptation that explicitly conditions on predicted weather unlocks robust street-view perception even with non-street-view training data.
Adverse conditions such as rain, snow, fog, and dust remain challenging for camera-based perception in autonomous driving. We study multi-class weather recognition from street-view images under domain shift, where most available training data come from non-street-view sources that differ markedly from real driving scenes. We propose Weather-Aware Adversarial Discriminative Domain Adaptation (WA-ADDA), which conditions the domain discriminator on predicted weather to promote features that are both domain-invariant and weather-sensitive. We also assemble a multi-dataset benchmark by unifying diverse non-street-view weather collections as sources and real street-view images as targets, and define a standardized evaluation protocol with macro accuracy as the primary metric. Across backbones (ResNet-50, EfficientNet, VGG, DenseNet), WA-ADDA consistently improves street-view performance and yields strong per-class recalls in challenging conditions while preserving clear-weather accuracy. These findings highlight the feasibility of domain-adapted weather recognition and the value of our benchmark for advancing robust, on-board perception.