turtlezhong / hfnet_ros Goto Github PK

I run this project in docker,but it can not get result.

Thanks for your great job, i am confused that how to match two descriptor? Is there some c++ interface?

Thank you for your work. I am running on NVIDIA GeForce GTX 1060, but the frame rate is limited to 6Hz. I would like to know if there is a way to speed it up. I am considering using it on the Jetson platform later.

I am using images of size 240x320 and have reduced the number of feature points extracted to 300, which seems to have no impact on the processing time.

// topic /debug_pts HZ
average rate: 6.305
min: 0.143s max: 0.792s std dev: 0.02366s window: 8032
average rate: 6.305
min: 0.143s max: 0.792s std dev: 0.02365s window: 8038
average rate: 6.305
min: 0.143s max: 0.792s std dev: 0.02364s window: 8044
average rate: 6.305

// program output
detected: 317 pts
IMAGE[240 x 320] cost: 159 ms
detected: 328 pts
IMAGE[240 x 320] cost: 158 ms
detected: 325 pts
IMAGE[240 x 320] cost: 160 ms
detected: 327 pts
IMAGE[240 x 320] cost: 159 ms
detected: 322 pts
IMAGE[240 x 320] cost: 161 ms
detected: 323 pts
IMAGE[240 x 320] cost: 160 ms
detected: 318 pts
IMAGE[240 x 320] cost: 160 ms

Thank you for your great work! I would like to ask if this project has completed the quantification and acceleration of the original algorithm hfnet