RESA: Recurrent Feature-Shift Aggregator for Lane Detection

Abstract

Lane detection is one of the most important tasks in self-driving. Due to various complex scenarios (e.g., severe occlusion, ambiguous lanes, etc.) and the sparse supervisory signals inherent in lane annotations, lane detection task is still challenging. Thus, it is difficult for the ordinary convolutional neural network (CNN) to train in general scenes to catch subtle lane feature from the raw image. In this paper, we present a novel module named REcurrent Feature-Shift Aggregator (RESA) to enrich lane feature after preliminary feature extraction with an ordinary CNN. RESA takes advantage of strong shape priors of lanes and captures spatial relationships of pixels across rows and columns. It shifts sliced feature map recurrently in vertical and horizontal directions and enables each pixel to gather global information. RESA can conjecture lanes accurately in challenging scenarios with weak appearance clues by aggregating sliced feature map. Moreover, we propose a Bilateral Up-Sampling Decoder that combines coarse-grained and fine-detailed features in the up-sampling stage. It can recover the low-resolution feature map into pixel-wise prediction meticulously. Our method achieves state-of-the-art results on two popular lane detection benchmarks (CULane and Tusimple). Code has been made available at: https://github.com/ZJULearning/resa.

Research on Lane Line Detection Algorithm Based on Instance Segmentation

“…It is also simple to ignore coarse-grained features and other issues. To solve the above problems, Zheng et al [ 24 ] proposed a bilateral upsampling module, which directly adds bilinear interpolation and transposed convolution upsampling results, and achieved certain results, but did not consider the two kinds of upsampling based on the applicability of the method to a specific image area. To effectively extract image features, this paper suggests an adaptive upsampling module that enables the network to choose the weight of the two upsampling methods at each location.…”

BEV Lane Det: Fast Lane Detection on BEV Ground

“…In the 2D lane detection, it is divided into three directions, pixel-wise segmentation, row-wise segmentation, and curve parameters. [25,34] consider 2D lane detection as a multi-category segmentation task based on pixel-wise, these methods set the limit number of lane lines, and the computing cost is expensive. 2D lane detection is also regarded as two-class segmentation by [18,22], and then combines with the embedding way to cluster each lane line to achieve the variable number of lane detection.…”

Development of an Autonomous Driving Vehicle for Garbage Collection in Residential Areas

“…CNN in Yu et al [ 6 ] was constructed by adding a perspective transform layer that enables accurate semantic segmentation of the lane, even when the pixels are reduced according to the distance of the lane seen in the image. Zheng et al [ 5 ] showed a recurrent feature-shift aggregator (RESA) to acquire enriched lane features from general CNN features using the spatial relationships of pixels across rows and columns. Hou et al [ 7 ] applied the self-attention distillation (SAD) method to perform lane detection by improving ENet [ 2 ] performance without any additional data or labels.…”

“…Recently, with the rapid development of deep learning, the technology for recognizing the surrounding environment through sensor data has also developed significantly. These deep learning technologies are commonly used for recognizing lanes through cameras [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 ] or for recognizing surrounding objects [ 11 , 12 , 13 , 14 ], or as a method for simultaneously recognizing surrounding objects using the camera and lidar [ 15 , 16 ] in autonomous vehicles. Along with various high-precision sensors and deep learning technologies described above, research and commercialization of autonomous vehicles that can autonomously drive on roads without human intervention are also rapidly progressing.…”