search

marco-rudolph / cs-flow

This is the official repository to the WACV 2022 paper "Fully Convolutional Cross-Scale-Flows for Image-based Defect Detection" by Marco Rudolph, Tom Wehrbein, Bodo Rosenhahn and Bastian Wandt.
111 stars 30 forks source link

About the detection time #3

Closed xfby2016 closed 2 years ago

xfby2016 commented 2 years ago

Hi,when I set pre_extracted = False, running evaluate.py, the calculation time for each image was approximately 0.4 seconds(use cuda),Is there any way to improve the performance? thanks !

marco-rudolph commented 2 years ago

Hi,

as long as feature extraction is the bottleneck here with EfficientNet as the feature extractor, you could change the feature extractor (e. g. to AlexNet), select an earlier layer or reduce image size.

Best regards Marco Rudolph