search

ijkguo / mx-rcnn

Parallel Faster R-CNN implementation with MXNet.
Other
669 stars 292 forks source link

cannot reproduce faster rcnn mAP #72

Open jiangxiaoyan opened 7 years ago

jiangxiaoyan commented 7 years ago

i run bash script/vgg_voc07.sh 1,2,3, use three GPU cards， Tesla M40 but "root:Means AP=0.6860", not like 70.23 image can you give the detail for experiment config

also, i test 2 GPU cards. the mAP is only 0.6303. i do not understand why this.

mAtthEwwww commented 7 years ago

i run bash script/resnet_voc0712.sh 1, 2 but only got mAP 71%, far less than 79% @precedenceguo

ijkguo commented 7 years ago

Please check if today's commit from mxnet pr #6849 fixed the problem.

Zehaos commented 6 years ago

@jiangxiaoyan @mAtthEwwww Can you reproduce the result now? Thanks.

mAtthEwwww commented 6 years ago

@Zehaos i have gave up

jonbakerfish commented 6 years ago

I can't reproduce the resnet101+voc2017+voc2012 either. I'm using MXNet 0.11.1. The mAP I got is ~0.69.

wassryan commented 6 years ago

I met the same problem as @ @jiangxiaoyan @mAtthEwwww ,I got vgg+alternate by 68.3MAP.Another problem is that I train resnet101+alternate,I got the high train accuracy and low loss,but only got a low test accuracy which is only 28MAP.What happened to me,anyone help?

ijkguo commented 6 years ago

I find VGG still works by these steps. Results can fluctuate in 68 to 70 between experiments. Usually it is 69.xx or 70.xx. Will look into resnet.

ijkguo commented 6 years ago

We did not fix random seed so results would vary between experiments. However, it is always around the reference on my end. Did you try to evaluate the released models?

315386775 commented 5 years ago

@k-miracle i have the same problem. I train resnet101+end2end, I got the high train accuracy and low loss,but only got a low test accuracy and map. img_pixel_means = (0.0, 0.0, 0.0).

ijkguo commented 5 years ago

Please evaluate the released model. Let's look at inference stage first.