Closed jcyhcs closed 3 years ago
I'm working now on researching and developing a lighter version of network that is based on the SuperPoint approach, to make it suitable for embedded devices. The development is going in the axial-attention-backbone. Hope to finish soon.
hi,professor: when run superpoint algorithm on jetson embedded platform, use pytorch or libtorch is so heavy, they rely on so huge thrid-party libraries, can you realize this algorithm use native cuda and tensorRT ? that will be greatfull! PLEASE HELP!