search

SJTU-ViSYS / FeatureBooster

FeatureBooster: Boosting Feature Descriptors with a Lightweight Neural Network (CVPR 2023)
Apache License 2.0
277 stars 16 forks source link

Demo code #2

Closed TuanTNG closed 1 year ago

TuanTNG commented 1 year ago

Hi, Thank you for your excellent work.

Can you provide a Python script that can run inference from 2 images? By the way, have you tested the end-2-end inference speed?

Looking forward to hearing from you soon.

Antu3heng commented 1 year ago

Hi. Sorry for the late response.

Can you provide a Python script that can run inference from 2 images?

For the demo code, you can refer to our Jupyter Notebook scripts in qualitative/qualitative-matches.ipynb. It provides a example for matching a pair of images using different features.

have you tested the end-2-end inference speed?

It depends on the number of features. We mainly tested the runtime of ALIKE-L on Jetson Xavier NX. It took about 125ms for the extraction of ALIKE, and our FeatureBooster took extra 13/23/46ms to boost 500/1000/2000 ALIKEs.

Besides, by using our Jupyter Notebook scripts in qualitative/qualitative-matches.ipynb, you can easily to tested the end-to-end inference speed by yourself.