search

tomrunia / TF_FeatureExtraction

Convenient wrapper for TensorFlow feature extraction from pre-trained models using tf.contrib.slim
180 stars 61 forks source link

is it possiple to use this code to extract the feature with my own trained model #6

Open jren2017 opened 7 years ago

jren2017 commented 7 years ago

I trained a model on my own dataset, is it possible to use this code to extract the feature?

Lucky94 commented 7 years ago

HI @jren2017, It is possible but will require a lot of changes and even will require you to put your model source code inside slim and link it everwhere in the code carefully, I have not tried that but I hope it should work.

Instead I would sugest you to go through this script from @davidsandberg https://github.com/davidsandberg/facenet/blob/master/contributed/export_embeddings.py, just modify it according to your needs, it will work.

Ps: thanks to @davidsandberg

Hope it helps. Regards, Lucky Srivastava

bkwwangjie commented 6 years ago

Hello, I would like to ask if I use the data set in my own, I finally extract the matrix to be saved to the txt file instead of HDF5 file, the extraction format is: the name of the file in the file + space + feature matrix, how do I write Code? Thank you~

bkwwangjie commented 6 years ago

@tomrunia Look forward to your reply to me。

saira05 commented 6 years ago

i got this error how to resolve it?

File "E:\TF_FeatureExtraction-master\feature_extractor\feature_extractor.py", line 20, in from nets import nets_factory ModuleNotFoundError: No module named 'nets'

hannarud commented 6 years ago

@Saira05 - you probably forgot to add /models/research/slim to you PYTHONPATH, see README setup 2nd step https://github.com/tomrunia/TF_FeatureExtraction#setup