I'd like to ask what the detection threshold boundaries are (min and max)? I assumed it was bounded between 0 and 1, but I've had cases where np.max(textmap[labels == component_id]) is greater than 1, and probabilities/confidence scores shouldn't exceed 1!
What do the values in textmap represent, since my initial assumption that the values are probabilities/confidence scores seems to be wrong.
Note: In the original CRAFT implementation, this greater than 1 occurence was also observed when I tested it.
Hi,
I'd like to ask what the detection threshold boundaries are (min and max)? I assumed it was bounded between 0 and 1, but I've had cases where np.max(textmap[labels == component_id]) is greater than 1, and probabilities/confidence scores shouldn't exceed 1!
What do the values in textmap represent, since my initial assumption that the values are probabilities/confidence scores seems to be wrong.
Note: In the original CRAFT implementation, this greater than 1 occurence was also observed when I tested it.
Thank you very much for your attention.