I have some images that I first identified objects inside the images with the Object Detection API. But the detected objects inside each bounding boxes are either rotated or