CrowdCounting-P2PNet icon indicating copy to clipboard operation
CrowdCounting-P2PNet copied to clipboard

Published 20 hours ago • verified publisher icon TencentYoutuResearch

精度差

Open FengJin-cv opened this issue 3 years ago • 3 comments

你好，请问我使用ShanghaiTech_Crowd_Counting_Dataset/part_A_final数据进行训练，参数使用的是readme里面的参数，但是精度很差，是什么原因？ 2779][lr 0.0001000][11.72s]mae:329.14285714285717, mse:466.4266075383996, time:10.338678121566772, best mae:295.75274725274727metric/mae@2778: 329.14285714285717metric/mse@2778: 466.4266075383996loss/loss@2780: 0.04982551398712236loss/loss_ce@2780: 0.04982551398712236[ep 2780][lr 0.0001000][12.00s]

FengJin-cv avatar Jan 25 '22 05:01 FengJin-cv

已解决。

FengJin-cv avatar Jan 25 '22 06:01 FengJin-cv

已解决。

您好，我也遇到相似的问题了，请问您是怎么解决的呢？

Ssssssyz avatar Mar 02 '22 00:03 Ssssssyz

你好，为训练的过程会出现显存爆炸，哪怕设置bachsize=1都如此，请问你怎么解决的？

Tomhardy13 avatar Dec 15 '22 09:12 Tomhardy13