Project translated automatically. Log in or register, to view the original The task There is a picture In him Picture 1 Picture 2 I need to compare image 1 with image 2 Showing a coefficient similar All photos are color inventions of products on white. In couples compared are identical and completely unlike images. But even the photographs of identical goods can differ both by the size of the object itself and by the size of the white fields around the product. That is, you need to first drop the white background before comparison, and then bring the comparative pair to the same size. There is a base of approximately 10k pairs of already compared manual goods.Compare hands, accordingly, the majority ok, but there are errors. As I understand, on this basis you can learn the model. Python scripts must be called from the command line, and get as parameters the path to the photos Now the photo comparison is realized in this way. from skimage.metrics import structural_similarity Imported cv2 Imports by SYS print(sys.argv ) print(sys.argv[2]) image1 = cv2.imread(sys.argv[1]) image2 = cv2.imread(sys.argv[2]) image1 = cv2.cvtColor(image1, cv2.COLOR_BGR2GRAY) image2 = cv2.cvtColor(image2, cv2.COLOR_BGR2GRAY) similarity = structural_similarity(image1, image2) print('result~~~', similarity) It does not always give logical results.

Proposals 4 Discussions 1