Arbitrary Shape Text Detection via Boundary Transformer；The paper at: https://arxiv.org/abs/2205.05320, which has been accepted by IEEE Transactions on Multimedia (T-MM 2023).
Hello, I noticed that there is an inaccuracy in the TextNet class in method forward the test_size is the same for h and w.
Also, for compatibility with cv2, I added a method LooseVersion to check the current version.
Since I got errors that the contours were not found.
Still correct indicated more specific data types for numpy
Hello, I noticed that there is an inaccuracy in the TextNet class in method forward the test_size is the same for h and w. Also, for compatibility with cv2, I added a method LooseVersion to check the current version. Since I got errors that the contours were not found. Still correct indicated more specific data types for numpy