xuannianz / keras-GaussianYOLOv3

Gaussian YOLOv3 (An Accurate and Fast Object Detector Using Localization Uncertainty for Autonomous Driving (ICCV, 2019)) implementation in Keras and Tensorflow

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

guide to pytorch implement

harlem867 opened this issue · comments

@xuannianz Hello, i am trying to implement guassian yolov3 in pytorch,
and i am using your guassian loss function.
During training, the guassian loss of x,y,w,h can reduce as normal, but in the test, all the test mAPs are 0. So i checked the training process, and have found that the mean of x,y,w,h didn't approach to the true value of GT while the sigmas can be rightly reduced to zero.
Can you give some advice? thanks a lot :)

Hi, how do you find the means do not approach to the gt values? Did the losses about x,y,w,h decrease?

1)yes,the gaussian loss of all the x,y,w,h can be rightly reduced. But while testing, the mAP is 0.
2)So I checked the training process, caculate and print the MSE between x and mean_x(the true value) and the MSE between sigma and 0, and find the means do not approach to the gt values, so the predicted result is not right.

commented

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.