Joshua Z. Zhang

use a vpn and try again?

@123chengbo Here you go http://pan.baidu.com/s/1pKVIdyN 4tjb

Reduced? Did you mean finetune?

checkout `--finetune` option to use pretrained models.

Did you download the pretrained model?

are you using large enough data shape, say 512?

Well, if you are trying to write a iterator for segmentation, then all these classes I wrote will be necessarily rewritten to have full functionalities. io/iter_image_det_recordio.cc: ImageDetLabelMap: used to parse...

cuda and cudnn version? The seg fault is very likely a library linking problem due to newer version of cuda or cudnn I guess.

I've checked the log again, since the detection time is printed out, the detection is finished, so I think that would be related to display or drawing problem, i.e. matplotlib...

Use the latest mxnet, the problem should be resolved. You can tune the parameters to make the crop more/less if you wish.