Approach for Lane Detection Using Key Points Estimation and Point Instance Segmentation PROJECT TITLE : Key Points Estimation and Point Instance Segmentation Approach for Lane Detection ABSTRACT: Techniques of perception used in autonomous vehicles should be adaptable to the various environments they encounter. Essential perception modules for traffic line detection should take into account a variety of factors, including the total number of traffic lines and the computing capability of the target system, among other conditions. In this paper, we propose a traffic line detection method that we call Point Instance Network (PINet). The method is founded on the key points estimation and instance segmentation approach. The purpose of this method is to address the issues that have been raised. The PINet makes use of a number of different hourglass models, all of which undergo simultaneous training with the same loss function. Because of this, the size of the trained models can be adjusted according to the amount of computing power available in the target environment. We recast the clustering problem of the predicted key points as an instance segmentation problem; this allows for the PINet to be trained regardless of the number of traffic lines. On the well-known public datasets CULane and TuSimple, both of which are used for lane detection, the PINet achieves competitive accuracy and false positive rates. Did you like this research project? To get this research project Guidelines, Training and Code... Click Here facebook twitter google+ linkedin stumble pinterest Mining Deep Model Data Impressions to Replace the Lack of Training Data Social media traffic data mining using the MC-LSTM-Conv model