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.
? The CNN models lowcomplexity local patterns with two separate heads, the first one predicts the existence of key points, and the second refines the location of key points in the local range and correlates key points of the same lane line.
? Most existing methods adopt well-studied frameworks such as semantic segmentation and object detection to parse lane markers and transform the network output into parametric curves through post-processing.
? The heatmap outputted by the first head expresses the possibility that keypoint appears, which resolves the existence of local curves.
? A specific feature map was predicted to indicate the position of a lane marker on each row.
? 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 within the lane class.
? In particular, the neural network takes as input the image and is optimized with a loss function that is tailored to the lane fitting problem.
? A branched, multi-task architecture to cast the lane detection problem as an instance segmentation task, that handles lane changes and allows the inference of an arbitrary number of lanes.
? We train a neural network end-to-end for lane detection, in a way that copes with the aforementioned problem of lane switching as well as the limitations on the number of lanes.
• In this work, we propose CondLaneNet, a novel top-to-down lane detection framework that detects the lane instances first and then dynamically predicts the line shape for each instance.
• We have greatly improved the ability of lane instancelevel discrimination by the proposed conditional lane detection strategy and row-wise formulation.
• To overcome this problem, anchor-based methods and row-wise detection methods were proposed.
• CurveLanes is a recently proposed benchmark with cases of complex topologies such as fork lines and dense lines.
• Based on the local information, two decoding algorithms with different preferences are proposed to predict global geometry of lane markers.
? To increase performance, both in terms of speed and accuracy, these two parts are jointly trained in a multi-task network.
? To increase the quality of the fit while retaining computational efficiency, it is common to convert the image into a ”bird’s-eye view” using a perspective transformation and perform the curve fitting there.
? To remedy this situation we also apply a perspective transformation onto the image before fitting a curve, but in contrast to existing methods that rely on a fixed transformation matrix for doing the perspective transformation, we train a neural network to output the transformation coefficients.
? An inherent advantage of the proposed method is that the lane fitting is robust against road plane changes and is specifically optimized for better fitting the lanes.
|