The project focusses on oriented weapons detection with a state of the art dataset. We collected a dataset which has more real life scenarios with blur images of weapons on all angles from 0 to 180 degree and up to 23 instances per image in order to generalize the oriented object detection models well.