We propose an image based end-to-end learning framework that helpslane-change decisions for human drivers and autonomous vehicles. The proposedsystem, Safe Lane-Change Aid Network (SLCAN), trains a deep convolutionalneural network to classify the status of adjacent lanes from rear view imagesacquired by cameras mounted on both sides of the vehicle. Rather than dependingon any explicit object detection or tracking scheme, SLCAN reads the wholeinput image and directly decides whether initiation of the lane-change at themoment is safe or not. We collected and annotated 77,273 rear side view imagesto train and test SLCAN. Experimental results show that the proposed frameworkachieves 96.98% classification accuracy although the test images are fromunseen roadways. We also visualize the saliency map to understand which part ofimage SLCAN looks at for correct decisions.
展开▼
