search

ijkguo / mx-rcnn

Parallel Faster R-CNN implementation with MXNet.
Other
671 stars 290 forks source link

How can I train the data without the pretrained model ? #52

Closed mu001999 closed 7 years ago

mu001999 commented 7 years ago

1, I can't download the vgg-0000.params, because it not appeared when I open the download link. 2, If I can download the params, I also need to train my data without the pretrained params. 3, When I make the code about arg_params and aux_params between '#' or ''', it will come with the error: out of memory.

I need help, thanks very much !

mu001999 commented 7 years ago

@precedenceguo

ijkguo commented 7 years ago

You cannot.

Oh, download link is broken. see https://github.com/dmlc/mxnet-model-gallery/blob/master/imagenet-1k-vgg.md

mu001999 commented 7 years ago

Thanks very much, but could you tell me why cannot ?

ijkguo commented 7 years ago

With the imagenet pretrained model, we don't need to randomly initialize the conv layers. So I did not write the initialization code. So you cannot unless you do it yourself.