KPNDepth: Depth Estimation of Lane Images under Complex Rainy Environment
This work addresses a domain-specific problem for autonomous driving systems by improving depth estimation under adverse weather, though it is incremental as it builds on existing methods for clear conditions.
The paper tackles depth estimation in rainy lane images by introducing a dual-layer convolutional kernel prediction network and a synthetic dataset, achieving effective performance in complex rainy conditions.
Recent advancements in deep neural networks have improved depth estimation in clear, daytime driving scenarios. However, existing methods struggle with rainy conditions due to rain streaks and fog, which distort depth estimation. This paper introduces a novel dual-layer convolutional kernel prediction network for lane depth estimation in rainy environments. It predicts two sets of kernels to mitigate depth loss and rain streak artifacts. To address the scarcity of real rainy lane data, an image synthesis algorithm, RCFLane, is presented, creating a synthetic dataset called RainKITTI. Experiments show the framework's effectiveness in complex rainy conditions.