LaneAF: Robust Multi-Lane Detection with Affinity Fields

Abstract

This study presents an approach to lane detection involving the prediction of binary segmentation masks and perpixel affinity fields. These affinity fields, along with the binary masks, can then be used to cluster lane pixels horizontally and vertically into corresponding lane instances in a post-processing step. This clustering is achieved through a simple row-by-row decoding process with little overhead; such an approach allows LaneAF to detect a variable number of lanes without assuming a fixed or maximum number of lanes. Moreover, this form of clustering is more interpretable in comparison to previous visual clustering approaches, and can be analyzed to identify and correct sources of error. Qualitative and quantitative results obtained on popular lane detection datasets demonstrate the model's ability to detect and cluster lanes effectively and robustly. Our proposed approach performs on par with stateof-the-art approaches on the limited TuSimple benchmark, and sets a new state-of-the-art on the challenging CULane dataset.

Accurate and Lightweight RailNet for Real-Time Rail Line Detection

“…Next, In the process of feature point detection in a sliding window, the window pixels are traversed, and the coordinates of non-zero pixel values are recorded. When the number of effective pixels in the window is less than the threshold, the window width is increased by the window height and width until the minimum number of pixels is met [29]. Furthermore, taking the average value of the abscissa of the effective pixels in the sliding window as the base point coordinate of the next sliding window, iterative detection is carried out until the total number of sliding windows is satisfied [30].…”

“…Secondly, we label the rail lines of all the rail images by using LABELME to get the JSON file used as the real rail lines during training and finally compared with the predicted rail lines [29]. In the actual training, the 3000 pictures are divided into the training set, verification set, and test set according to the ratio of 0.9:0.05:0.05.…”

K-Lane: Lidar Lane Dataset and Benchmark for Urban Roads and Highways

“…Compared with the early rule-based techniques (Dong et al 2012;Deng and Wu 2018), CNNbased lane detection techniques are more adaptive to various weather changes and show less performance deterioration by occlusion. In these techniques, lanes are predicted by a lane detection head based on local features extracted by CNN (He et al 2016), and the performance is improved with development of lane detection heads that exploit the features of lane lines; Segmentation-based techniques such as (Pan et al 2018) detect lanes by assigning classes (e.g., lanes, and backgrounds) to each predicted pixel, which may cause discontinuous lane lines and marks and additional clustering is introduced to compensate (Abualsaud et al 2021). Anchorbased techniques detect lanes through the regression of coordinate change of anchors that are designated initially.…”

“…To design the the detection head, we formulize the lane detection problem as a multi-class segmentation problem, where each pixel is assigned a class and a confidence score. This is because there are multiple advantages with the multiclass segmentation-based approaches over other approaches; First, binary-class (i.e., lane or background) segmentationbased approaches (Abualsaud et al 2021) need additional post-processing to assign a new lane class for a new prediction, while multi-class segmentation-based approaches assign a class directly to a prediction. Second, anchorbased approaches use fixed-shaped anchors that limits the lane shape detection.…”

Lane Detection with Versatile AtrousFormer and Local Semantic Guidance

“…As for those anchor-free, they equate lane detection to high-order polynomial regression, straightforward yet overly relying on certain parameters. The last main group [23,24,25,26], inspired by the fancy thoughts from human pose estimation, usually extract key points with lane semantic and then cluster them into different lane instances via complex post-processing methods. In general, methods other than sementic segmentation have difficulties in modeling more complex lane forms, like those described in BDD100K [27].…”