movchan74/street_to_shop_experiments

street_to_shop_experiments's Issues

which version of keras and python do you use?

I have tried to load the keras model use python2.7 and python3.6, but received an error like this:

  File "/opt/miniconda3/lib/python3.6/site-packages/keras/layers/core.py", line 663, in call
    return self.function(inputs, **arguments)
  File "<ipython-input-8-be15121d8712>", line 7, in <lambda>
SystemError: unknown opcode

Is this error caused by the Lamda function? Could you please show me the version of python, keras and tensorflow in this project? thanks a lot!

get_item_info

how to call the function get_item_info? eg. how get the product list?
could you share this?

to improve the result

hi Movchan
thanks your code and blog first. it is so cool. after that do you make any other effort to improve the result ? as you know. the user upload the image with noise. we should focus the main object. add a main part object detect net? or a attention net base the cnn ? I also meet the problem about the domain adaption. I have no good idea. Besides i think the triplet loss based on the cosin or angular will be better