Detection of Lanes and Object for Self Driving Vehicles using Deep Learning and Flask

Abstract

Self-driving cars are crucial for decreasing Accidents on the road, traffic congestion, energy conservation, and environmental concerns preservation, among other things. An autonomous driving perception system's basic purpose is real-time lane detection. Lane detection is currently reliant on image processing as the primary method, and there are other detection algorithms. However, because the algorithm model is complicated, and the computational volume is big. The overwhelming majority of it can only be implemented on the GPU+GPU platform, resulting in low-cost performance, high power consumption, and a large bulk that is incompatible with vehicle-mounted requirements. To satisfy real-time lane detecting requirements. Faster R-CNN is extended by Object Detection, which adds In addition to the present branch for bounding box identification, There is a branch for anticipating the mask of an item. Mask R-CNN is easy to use and adds just a little amount of overhead to Faster R-CNN, which runs at 5 frames per second. Mask RCNN may also be used for a variety of other tasks, such as assessing human postures. within a similar framework In all three COCO challenge tracks, we get great results: object detection, person keypoint detection.