Hi,
I am trying to use fastcv decriptors for object recognition.
According to my experiments, the 36 bytes descriptor seems to be a 3x3 oriented matrix of 4 bytes.
What is the best way to make the 36 bytes descriptor robust against scaling and rotation, like SIFT descriptors?
Cyril
Hey ciryl, could you share the code you used to create the 17x17 patch and the descriptors?
I'm also trying to use fastcv for object recognition (see the post below yours).
Thank you! :)
Sincerly, Steve.