The function of mask_ Minimumweightmatching may have time consuming problem

megvii-research / Iter-E2EDET

Official implementation of the paper "Progressive End-to-End Object Detection in Crowded Scenes"

MIT License

88 stars 8 forks source link

The function of mask_ Minimumweightmatching may have time consuming problem #6

Closed BubblyYi closed 2 years ago

BubblyYi commented 2 years ago

Hi author! Thanks for your amazing works

When I run the training code, I find that the training speed is extremely slow, and finally locating the function of mask_ Minimumweightmatching() takes up most of the time. thus step_forward() function is compared with the regular_forward() will be 20 times slower. It may take about 1 hour to train 100 pictures(training on GPU has been checked). Is this normal?

Waiting for your reply! James

BubblyYi commented 2 years ago

It is a low scipy version problem, please use the version that >=1.5.2