Towards End-to-End Lane Detection: an Instance Segmentation Approach

Abstract

Modern cars are incorporating an increasing number of driver assist features, among which automatic lane keeping. The latter allows the car to properly position itself within the road lanes, which is also crucial for any subsequent lane departure or trajectory planning decision in fully autonomous cars. Traditional lane detection methods rely on a combination of highly-specialized, hand-crafted features and heuristics, usually followed by post-processing techniques, that are computationally expensive and prone to scalability due to road scene variations. More recent approaches leverage deep learning models, trained for pixel-wise lane segmentation, even when no markings are present in the image due to their big receptive field. Despite their advantages, these methods are limited to detecting a pre-defined, fixed number of lanes, e.g. ego-lanes, and can not cope with lane changes. In this paper, we go beyond the aforementioned limitations and propose to cast the lane detection problem as an instance segmentation problem -in which each lane forms its own instance -that can be trained end-to-end. To parametrize the segmented lane instances before fitting the lane, we further propose to apply a learned perspective transformation, conditioned on the image, in contrast to a fixed "bird's-eye view" transformation. By doing so, we ensure a lane fitting which is robust against road plane changes, unlike existing approaches that rely on a fixed, predefined transformation. In summary, we propose a fast lane detection algorithm, running at 50 fps, which can handle a variable number of lanes and cope with lane changes. We verify our method on the tuSimple dataset and achieve competitive results.

Robust Lane Detection From Continuous Driving Scenes Using Deep Neural Networks

“…The end-to-end encoder-decoder network is built on the basis of road scene object segmentation task, trained on ImageNet. In [39], a network called LaneNet was proposed. LaneNet is built on SegNet, but has two decoders.…”

Enhanced free space detection in multiple lanes based on single CNN with scene identification

“…1 rely on hand-crafted kernels, but with an increasing need for labeled datasets and computing power. The CNNs for lane detection are usually trained to detect lane boundaries [6] or lane markings [7]. Lastly, the output of the network is post-processed and filtered, to get information about the geometric structure of the lanes.…”

“…Here, some approaches use a clustering algorithm to distinguish between different lanes [8]; others fit a model for each segment of interest, usually a polynome. For instance, Neven et al [6] train a CNN to separate the pixels belonging to different lane boundaries, then they fit the detected points using a third-degree polynome. Many others ( [9], [10]) use the RANSAC algorithm to remove outliers.…”

Multi-lane Detection Using Instance Segmentation and Attentive Voting

“…It can be exploited by limiting the depth of the encoder, by increasing the inference speed and by decreasing learnable parameters in lane segmentation network. Apart from lane segmentation, the state-of-the-art method like [12] strongly depends on the road scene environment to generate the embeddings for clustering. Intuitively, using lane markings alone for multi-lane detection makes more sense.…”