Improving Nighttime Driving-Scene Segmentation via Dual Image-adaptive Learnable Filters

Semantic segmentation on driving-scene images is vital for autonomous driving. Although encouraging performance has been achieved daytime images, the nighttime are less satisfactory due to insufficient exposure and lack of labeled data. To address these issues, we present an add-on module called dual image-adaptive learnable filters (DIAL-Filters) improve semantic in driving conditions, aiming at exploiting intrinsic features under different illuminations. DIAL-Filters consist two parts, including processing (IAPM) a guided filter (LGF). With DIAL-Filters, design both unsupervised supervised frameworks segmentation, which can be trained end-to-end manner. Specifically, IAPM consists small convolutional neural network with set differentiable image filters, where each adaptively enhanced better respect The LGF employed enhance output get final result. light-weight efficient they readily applied images. Our experiments show that DAIL-Filters significantly ACDC_Night NightCity datasets, while it demonstrates state-of-the-art Dark Zurich Nighttime Driving testbeds. Codes models available https://github.com/wenyyu/IA-Seg.