Multi-task processing method and device for intelligent driving

The embodiment of the invention provides a multi-task processing method and device for intelligent driving, a storage medium and electronic equipment. The multi-task processing method comprises the steps that 3D space grid voxels are generated based on a point cloud image; aiming at the 3D space grid voxels, extracting 3D high-dimensional features; acquiring an up-sampling feature based on the 3D high-dimensional feature; and obtaining road information based on the up-sampling features. According to the embodiment of the invention, the multi-task overall network is adopted, branch feature sharing of multiple tasks is realized through sharing the basic model, joint learning is beneficial to improving the performance of each task, and obstacle target detection, lane lines, travelable areas, ground heights and the like can be output at the same time; the integrated multi-task model is adopted to obtain better sensing performance and lower system delay at the same time, and the utilization rate of a deep learning