search

duanzhiihao / RAPiD

RAPiD: Rotation-Aware People Detection in Overhead Fisheye Images (CVPR 2020 Workshops)
http://vip.bu.edu/rapid/
Other
212 stars 63 forks source link

About validation time #3

Closed aia39 closed 4 years ago

aia39 commented 4 years ago

How much time does it take to validate on 1000 images? We are having some time complexities. We are not sure if there are any problems as we edited some part of your code.

duanzhiihao commented 4 years ago

Typically it costs around several minutes to validate on 1000 images, including prediction and metric calculation. The validation time is related to many factors, e.g., input image resolution and the network's precision.

If the network's precision is very low, i.e., it is not well trained, it will produce many garbage predictions that will significantly increase the metric calculation time. It would be great if you could share more detailed information, e.g., the input image resolution and how you train the network.