The Method of Classifying Fog Level of Outdoor Video Images Based on Convolutional Neural Networks

Unsupervised Contrastive Masking for Visual Haze Classification

“…Since it is hard to train models from complete haze images, related works improve the model training framework: model ensemble methods [22,27] achieve feature enhancement by training multiple basic models and adapting a meta-learner to learn to fusion basic models' output; the multi-branch method [17,28] proposes multiple classifiers in different training branches to get better predictions. There are also ways of pre-training [2,5], multitask training [32,33] and feature fusion [15,21]. The end-to-end method does not require feature engineering, however, most of the current works lack constraints on the background problem of haze classification.…”

“…Such data-driven approach alleviates introducing manual biases, but it introduces the need of feature engineering with high computational cost for feature extraction [2,22,28]. Finally, the deep learning algorithms use CNNs [32,33] with network ensembles [22,27], multi-branch training [17,28], and pre-training [2,5], to classify haze images in an end-to-end manner, which typically achieve much better performance than the threshold-based and the handcrafted-feature-based methods. However, it has been observed that these models usually perform worse on the classification of light haze images, due to the fact that the background scene objects work as spurious causal features in classification [16,27,30] and the lack of integration of corresponding features [6,12,20].…”