can't reach the mAP - Githubissues

chengyangfu / caffe

Caffe: a fast open framework for deep learning.

http://caffe.berkeleyvision.org/

Other

169 stars 47 forks source link

can't reach the mAP #16

Closed NickMinYang closed 6 years ago

NickMinYang commented 6 years ago

*It should reach 77.5 mAP at 80k iterations.** Why I can't get this map?I just clone the code, and follow the readme.

ghost commented 6 years ago

yes, me too, i just got 60+, how about you?

chengyangfu commented 6 years ago

The problem comes from the synchronization issue between Wei(The first author of SSD) and mine. However, right now I have no intention to use Caffe anymore. That means I will not fix current Github version. But I will release another version in PyTroch soon. and it will also fix some problems not mentioned in the paper.

rabbitwayne commented 5 years ago

@chengyangfu I have the same issue and mAP can only reach 0.68 during training. Can you say more about the synchronization issue? How to solve this problem in the current caffe release? When will you release the pytorch version of DSSD? Thanks!