Real-Time Fully Unsupervised Domain Adaptation for Lane Detection in Autonomous Driving
This enables real-time, unsupervised domain adaptation for lane detection in autonomous vehicles, though it is incremental as it builds on existing batch-normalization adaptation techniques.
The paper tackles the problem of adapting lane detection models to unseen environmental conditions in real-time for autonomous driving, achieving similar accuracy (avg. 92.19%) as a state-of-the-art semi-supervised method while operating at 30 FPS on Nvidia Jetson Orin.
While deep neural networks are being utilized heavily for autonomous driving, they need to be adapted to new unseen environmental conditions for which they were not trained. We focus on a safety critical application of lane detection, and propose a lightweight, fully unsupervised, real-time adaptation approach that only adapts the batch-normalization parameters of the model. We demonstrate that our technique can perform inference, followed by on-device adaptation, under a tight constraint of 30 FPS on Nvidia Jetson Orin. It shows similar accuracy (avg. of 92.19%) as a state-of-the-art semi-supervised adaptation algorithm but which does not support real-time adaptation.