Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

What's the error when I used the model.prototxt? #1

Open
getengqing opened this issue Jul 26, 2017 · 7 comments
Open

What's the error when I used the model.prototxt? #1

getengqing opened this issue Jul 26, 2017 · 7 comments

Comments

@getengqing
Copy link

I0726 09:27:07.791891 1097 net.cpp:98] Creating Layer data
F0726 09:27:07.791920 1097 net.cpp:394] Duplicate blobs produced by multiple sources.

@yjxiong
Copy link

yjxiong commented Jul 30, 2017

@innerlee
Any problem related to the conversion in Parrots2Caffe?

@shicai
Copy link

shicai commented Jul 30, 2017

@yjxiong hi, yuanjun, do you know how to evaluate this model?
I mean, readme.md missed some image preprocessing steps (mean/std values, scaled image size, etc).

@yjxiong
Copy link

yjxiong commented Jul 30, 2017

These models use standard Caffe-style preprocessing. That is 299 crops, 0-255 BGR images, and standard ImageNet mean values. You can consult here for the preprocessing steps

https://github.com/yjxiong/caffe/blob/action_recog/models/bvlc_googlenet/train_val.prototxt#L13

The only difference is that crop size should be 299.

@shicai
Copy link

shicai commented Jul 30, 2017

from model.prototxt, crop size should be 331.
so should i resize the input image to 378xN size?

@innerlee
Copy link
Collaborator

innerlee commented Jul 30, 2017

Hi shicai, for single crop evaluation:

  • resize to 378*378
  • crop to center 331*331
  • offset (104, 117, 123)

edit:
For multi-crop, top-k pooling is used. Will add all the details to the README.

@shicai
Copy link

shicai commented Jul 30, 2017

got it, thank you very much. @yjxiong @innerlee

@yjxiong
Copy link

yjxiong commented Jul 30, 2017

@innerlee

I suggest adding these to the README page.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants