I have a sequence of camera images overlapping each other, and I did an object detection by Faster RCNN deep network, which resulted in bounding boxes per image per each object, and I am going to do triangulation to convert bounding boxes vertices(2D pixel coordinates) to 3D coordinates in the world coordinate system. So, I need to do matching to find corresponding bounding boxes vertices over the images. Is it possible to use Sift matching to run "only" corresponding bounding boxes vertices in images overlapping? Moreover, is there any other way except matching to find corresponding bounding boxes over the images?
Below is a sample code that reproduces my method
#bounding box top left pixel coordinate for image 0 and image 1
bbox0 = [2337.07139369, 1132.20814732]
bbox1 = [2078.25994001, 1559.14941406]
#make key points for bbox0 and bbox1 using below code
key_points = cv2.KeyPoint(bbox[0], bbox[1], 1)]
#create a sift descriptor object
sift = cv2.SIFT_create()
#img0, and img1 RGB images read using cv2, and
#convert to gray images using cv2, #images shape : (4096, 3008)
#compute sift key points descriptors per each image
kp, des = sift.compute(image, key_points)
