Training method, device and equipment of lane-level positioning identification model and storage medium

The invention provides a training method, device and equipment for a lane-level positioning recognition model and a storage medium, and the method comprises the steps: constructing a simulation image sample and a marking image sample, enabling a high-frequency image to represent a lane line in the simulation image sample, and enabling the lane line to be marked in the marking image sample; constructing a neural network model by adopting a classification network, and generating a neural network model which takes the image as model input and takes the lane number as model output; performing neural network pre-training on the neural network model by adopting the simulation image sample to obtain a pre-training model; and carrying out neural network fine tuning training on the pre-training model by adopting the annotated image sample to generate a lane-level positioning identification model. Based on the method, the neural network is pre-trained by using the simulation image sample, and then the training network